Contemporary IMRT PDF

Series in Medical Physics and Biomedical Engineering
CONTEMPORARY IMRT
Developing Physics and Clinical Implementation
Steve Webb
Professor of Radiological Physics
Head, Joint Department of Physics
Institute of Cancer Research
University of London
UK
and
Royal Marsden NHS Foundation Trust
Sutton
Surrey
UK
Institute of Physics Publishing

Bristol and Philadelphia
Copyright 2005 IOP Publishing Ltd.
c IOP Publishing Ltd 2005

All rights reserved. No part of this publication may be reproduced, stored
in a retrieval system or transmitted in any form or by any means, electronic,
mechanical, photocopying, recording or otherwise, without the prior permission
of the publisher. Multiple copying is permitted in accordance with the terms
of licences issued by the Copyright Licensing Agency under the terms of its
agreement with Universities UK (UUK).
British Library Cataloguing-in-Publication Data
A catalogue record for this book is available from the British Library.
ISBN 0 7503 1004 9
Library of Congress Cataloging-in-Publication Data are available
Cover image: Multileaf Collimator, copyright Elekta, reproduced with
permission (see also figure 3.2, page 42).
Series Editors:
C G Orton, Karmanos Cancer Institute and Wayne State University, Detroit, USA
J H Nagel, Institute for Biomedical Engineering, University Stuttgart, Germany
J G Webster, University of Wisconsin-Madison, USA
Commissioning Editor: John Navas
Commissioning Assistant: Leah Fielding
Production Editor: Simon Laurenson
Production Control: Sarah Plenty
Cover Design: Victoria Le Billon
Marketing: Louise Higham and Ben Thomas
Published by Institute of Physics Publishing, wholly owned by The Institute of
Physics, London
Institute of Physics Publishing, Dirac House, Temple Back, Bristol BS1 6BE, UK
US Office: Institute of Physics Publishing, The Public Ledger Building, Suite
929, 150 South Independence Mall West, Philadelphia, PA 19106, USA
Typeset in LATEX 2 by Text 2 Text Limited, Torquay, Devon
Printed in the UK by MPG Books Ltd, Bodmin, Cornwall
The Series in Medical Physics and Biomedical Engineering is the official book
series of the International Federation for Medical and Biological Engineering
(IFMBE) and the International Organization for Medical Physics (IOMP).
IFMBE
The International Federation for Medical and Biological Engineering (IFMBE)
was established in 1959 to provide medical and biological engineering with a
vehicle for international collaboration in research and practice of the profession.
The Federation has a long history of encouraging and promoting international
cooperation and collaboration in the use of science and engineering for improving
health and quality of life.
The IFMBE is an organization with membership of national and transnational
societies and an International Academy. At present there are 48 national members
and two transnational members representing a total membership in excess of
30 000 world wide. An observer category is provided to give personal status
to groups or organizations considering formal affiliation. The International
Academy includes individuals who have been recognized by the IFMBE for their
outstanding contributions to biomedical engineering.
Objectives
The objectives of the International Federation for Medical and Biological
Engineering are scientific, technological, literary, and educational. Within the
field of medical, clinical and biological engineering its aims are to encourage
research and the application of knowledge, and to disseminate information and
promote collaboration.
In pursuit of these aims the Federation engages in the following activities:
sponsorship of national and international meetings, publication of official
journals, cooperation with other societies and organizations, appointment
of commissions on special problems, awarding of prizes and distinctions,
establishment of professional standards and ethics within the field, as well as other
activities which in the opinion of the General Assembly or the Administrative
Council would further the cause of medical, clinical or biological engineering.
It promotes the formation of regional, national, international or specialized
societies, groups or boards, the coordination of bibliographic or informational
services and the improvement of standards in terminology, equipment, methods
and safety practices, and the delivery of health care.
The Federation works to promote improved communication and understanding
in the world community of engineering, medicine and biology.
Activities
The IFMBE publishes the journal Medical and Biological Engineering and
Computing which includes a special section on Cellular Engineering. The IFMBE
News, published electronically, keeps the members informed of the developments
in the Federation. In cooperation with its regional conferences, IFMBE publishes

the series of IFMBE Proceedings. The Federation has two divisions: Clinical
Engineering and Technology Assessment in Health Care.
Every three years the IFMBE holds a World Congress on Medical Physics
and Biomedical Engineering, organized in cooperation with the IOMP and the
IUPESM. In addition, annual, milestone and regional conferences are organized
in different regions of the world, such as Asia Pacific, Baltic, Mediterranean,
Africa and South American regions.
The administrative council of the IFMBE meets once a year and is the
steering body for the IFMBE. The council is subject to the rulings of the General
Assembly, which meets every three years.
Information on the activities of the IFMBE are found on its web site at
http://www.ifmbe.org.
IOMP
The IOMP was founded in 1963. The membership includes 64 national societies,
two international organizations and 12 000 individuals. Membership of IOMP
consists of individual members of the Adhering National Organizations. Two
other forms of membership are available, namely Affiliated Regional Organization and Corporate members. The IOMP is administered by a Council, which
consists of delegates from each of the Adhering National Organizations; regular
meetings of council are held every three years at the International Conference
on Medical Physics (ICMP). The Officers of the Council are the President, the
Vice-President and the Secretary-General. IOMP committees include: developing countries, education and training; nominating; and publications.
Objectives
To organize international cooperation in medical physics in all its aspects,

especially in developing countries.
To encourage and advise on the formation of national organizations of
medical physics in those countries which lack such organizations.
Activities
Official publications of the IOMP are Physiological Measurement, Physics
in Medicine and Biology and the Series in Medical Physics and Biomedical
Engineering, all published by the Institute of Physics Publishing. The IOMP
publishes a bulletin Medical Physics World twice a year.
Two council meetings and one General Assembly are held every three years at
the ICMP. These conferences are normally held in collaboration with the IFMBE
to for the World Congress on Medical Physics and Biomedical Engineering. The
IOMP also sponsors occasional international conferences, workshops and courses.
Information on the activities of the IOMP are found on its web site at
http://www.iomp.org/.
Contents
Preface and acknowledgments
xv
Intensity-modulated radiation therapy (IMRT): General statements

and points of debate
1.1 Observations on IMRT at the current time
1.2 Criticism of the philosophy of IMRT
1
1
10
Developments in rotation IMRT and tomotherapy

2.1 NOMOS MIMiC tomotherapy
2.1.1 Technology history
2.1.2 Matchline concerns and solutions
2.1.3 Energy considerations
2.1.4 Concerns about increased treatment time
2.1.5 Machine features
2.2 University of Wisconsin machine for tomotherapy
2.2.1 Development history
2.2.2 Generation and use of megavoltage computed
tomography (MVCT) images
2.2.3 Clinical application
2.2.4 Commissioning issues
2.2.5 Verification of MIMiC and University of Wisconsin
tomotherapy
2.3 Tomotherapy using a 60 Co source
2.4 Tomotherapy with an MLC
2.5 Summary
18
18
18
21
23
24
25
26
26
Developments in IMRT using a multileaf collimator (MLC) (physics)

3.1 New sequencers/interpreters
3.1.1 General; dynamic IMRT sequencers with hard MLC
constraints on leaves and jaws
3.1.2 Sequencing multiple-static MLC fieldsclusters
3.1.3 Non-uniform spatial and fluence steps
3.1.4 Minimizing the number of segments
3.1.5 IMFAST
39
46
28
33
34
35
35
37
38
46
47
50
51
53
viii
Contents
3.1.6
3.1.7
3.1.8
The best interpreter ever?

53
Sequencers exploiting MLC rotation
56
Comparison of dynamic MLC (dMLC) and multiple
static field (MSF) techniques
58
3.1.9 Varian MLC and HELIOS planning system
61
3.1.10 Developments in Elekta IMRT
62
3.1.11 Other interpreters
63
3.1.12 The effect of removing the flattening filter
66
3.2 Radiation leakage and accounting for machine effects in IMRT
delivery
66
3.2.1 General leakage issues
66
3.2.2 Factoring in delivery physics
68
3.2.2.1 Measurement and prediction of leakage and
scatter
68
3.2.2.2 Using leakage and scatter knowledge in the
MSF-MLC technique
70
3.2.3 The effect of rounded leaf ends: light-field to
radiation-field discrepancy in IMRT
73
3.3 Dose calculation for IMRT
74
3.3.1 Application of colour theory
75
3.3.2 Penumbra sharpening for IMRT
76
3.3.3 MU verification
76
3.4 Features of MLC delivery of IMRT
77
3.4.1 Large-field IMRTsplitting the delivery
77
3.4.2 Tongue-and-groove effect
77
3.4.3 Leaf-speed limitations
80
3.4.4 Stability of accelerator and delivery of a small number of
MUs and small fieldsizes
81
3.5 Dynamic arc therapy
84
3.6 Combining step-and-shoot and dynamic delivery for dMLC
86
3.7 IMATtechnical issues
89
3.7.1 IMAT in clinical use
92
3.7.2 IMAT modified to aperture-modulated-arc therapy (AMAT) 94
3.8 New ideas related to the dMLC technique
95
3.9 Compensators and comparisons of compensator and MLC-based
IMRT
96
3.9.1 Do we need the MLC for IMRT?
96
3.9.2 Use of compensators for IMRT
97
3.9.3 Comparison of compensator and MLC-based IMRT
105
3.10 Optimum width of leaves for an MLC
106
3.11 MicroMLCs for IMRT
107
3.11.1 Siemens (virtual) microMLC
107
3.11.2 Varian (virtual) MLC
108
3.11.3 Elekta (virtual) microMLC and microMLC
108
Contents
3.12
3.13
3.14
3.15
4
ix
3.11.4 Radionics microMLC

109
3.11.5 BrainLAB microMLC
110
3.11.6 DKFZ-originating microMLCs
111
3.11.6.1 The MRC systems GMBH/Siemens Moduleaf 111
3.11.6.2 New DKFZ microMLC
112
3.11.7 3DLine microMLC
112
3.11.8 Comparison and use of microMLCs
112
3.11.9 Multi-level MLC
115
Increasing the spatial resolution of a conventional MLC
116
Verification of MLC-delivered IMRT
120
3.13.1 Electronic-portal-imager-based IMRT verification
120
3.13.2 Other EPID designs
124
3.13.3 Technical aspects of EPID imaging for IMRT
124
3.13.4 Extraction of anatomical images from portal images
generated during IMRT
125
3.13.5 Blocking-tray-level measurement
126
3.13.6 The two-level MLC
127
3.13.7 Water-beam-imaging system (WBIS)
129
3.13.8 Integrated portal fluence and portal dosimetry
131
3.13.9 IMRT verification phantom measurements
132
3.13.10 Verification by software techniques
138
3.13.11 Comparison of delivered modulated fluence profile with
plan-predicted modulated profile
141
3.13.12 Verification of canine and human IMRT using in-vivo
dosimetry
143
3.13.13 Polyacrylamide gel (PAG) dosimetry for IMRT verification 144
3.13.13.1 Review
144
3.13.13.2 PAG readout techniques
145
3.13.13.3 Use of PAGs for IMRT verification
147
3.13.13.4 New PAGs
150
3.13.14 Film as an IMRT dosimeter
150
Quality assurance (QA) of MLC delivery
152
3.14.1 Average leaf-pair opening (ALPO)
152
3.14.2 Routine QA of MLC leaf movement
153
3.14.3 Modelling the effects of MLC error
158
Summary
159
Developments in IMRT not using an MLC

4.1 The Cyberknife
4.2 The design of the shuttling MLC (SMLC)
4.3 IMRT with the jaws-plus-mask technique
4.4 The variable aperture collimator (VAC)
4.5 One-dimensional IMRT
4.6 Summary
161
161
168
169
174
176
177
x
5
Contents
Clinical IMRTevidence-based medicine?
5.1 IMRT of the prostate showing measurable clinical benefit
5.2 Comparison of treatment techniques for the prostate
5.3 Royal Marsden NHS Foundation Trust pelvic and other IMRT
5.4 Comparison of IMRT with conformal radiotherapy (CFRT) for
complex shaped tumours
5.5 IMRT for whole-pelvic and gynaecological radiotherapy
5.6 Head-and-neck IMRT
5.6.1 Thyroid IMRT
5.6.2 Nasopharynx IMRT
5.6.3 Oropharynx IMRT
5.6.4 Oropharynx and nasopharynx IMRT
5.6.5 Larynx IMRT
5.6.6 Evidence for parotid sparing
5.6.7 Meningioma IMRT
5.6.8 Paediatric medulloblastoma IMRT
5.6.9 Ethmoid cancer IMRT
5.6.10 Other studies reported
5.7 Breast IMRT
5.7.1 Breast IMRT at William Beaumont Hospital
5.7.2 Other reports of techniques using small top-up fields
5.7.3 EPID-based techniques for breast IMRT
5.7.4 Modified wedge technique
5.7.5 Breast IMRT in combination with use of respiration gating
5.7.6 Comparison of IMRT delivery techniques for the breast
5.7.7 Reduced complications observed following breast IMRT
5.7.8 Combination of IMRT with charged-particle irradiation
5.8 Bladder IMRT at the Christie Hospital
5.9 Lung cancer IMRT
5.10 Scalp IMRT
5.11 Other clinical IMRT reportsvarious tumour sites
5.12 Summary
3D planning for CFRT and IMRT: Developments in imaging for
planning and for assisting therapy
6.1 Challenges to IMRT and inverse planning
6.2 Determination of the GTV, CTV and PTV; the influence of 3D
medical imaging
6.2.1 General comments on inhomogeneous dose to the PTV
and image-guided planning
6.2.2 Interobserver variability in target-volume definition
6.2.3 Margin definition
6.2.4 Use of magnetic resonance for treatment planning
6.2.4.1 Distortion correction
179
183
187
192
197
199
201
202
203
204
205
205
206
209
210
210
210
211
211
214
216
218
218
218
220
222
222
224
225
227
228
230
230
231
231
231
232
234
234
Contents
6.2.4.2
6.2.4.3
6.2.4.4
6.2.4.5
6.2.4.6
6.3
6.4
Use of contrast agents

Planning based on MR images alone
Coregistered CT and MR planning
Increased protection of structures
Monitoring the response to radiotherapy via
MRI
6.2.5 Use of functional information from SPECT and PET for
treatment planning
6.2.5.1 Generalities and prostate imaging
6.2.5.2 Head-and-neck imaging
6.2.5.3 Lung imaging
6.2.5.4 Para-aortic lymph node (PALN) imaging
6.2.5.5 Combined PET-CT scanning
6.2.5.6 Imaging to overcome breathing-motion effects
6.2.6 Use of pathology specimens to compare with GTV and
PTV
New inverse-planning methods for IMRT
6.3.1 Gradient-descent inverse planning
6.3.2 Simulated annealing inverse planning
6.3.3 Equivalent-uniform-dose-based inverse planning
6.3.4 Maximum entropy inverse planning
6.3.5 Genetic algorithms
6.3.6 Single-step inverse planning
6.3.7 Simulated particle dynamics
6.3.8 Optimization of surrogate parameters in beam space
6.3.9 Comparison of inverse-planning techniques
6.3.10 Features and comparison of commercial planning
algorithms
6.3.11 Dependences of IMRT plans on target geometry
6.3.12 Multiple local minima and the global minimum in
optimization
6.3.13 Sampling the dose matrix for IMRT optimization speed-up
6.3.14 Creating a uniform PTV dose in IMRT; cost tuning
6.3.15 Importance factors
6.3.15.1 Voxel-dependent IFs
6.3.16 Biological and physical optimization
6.3.17 Pareto optimal IMRT
6.3.18 Combined CFRT and IMRT
6.3.19 Split modulation
6.3.20 Summary on inverse-planning techniques
New forward-planning methods for IMRT; direct aperture
optimization
6.4.1 Segmental inverse planning at Thomas Jefferson
University (TJU)
xi
236
237
237
240
241
242
242
244
246
247
248
248
250
250
251
251
251
252
253
254
255
257
259
261
262
263
266
269
269
271
273
275
275
276
278
278
279
xii
Contents
6.4.2
6.4.3
Aperture-based planning at the University of Ghent

Direct aperture optimization (DAO) at the University of
Maryland
6.4.4 DAO wobbling the MLC leaf positions
6.4.5 Other forward-planning studies
6.4.6 Summary on aperture-based IMRT
6.5 Smoothing IMBs
6.5.1 Smoothing techniques from the Royal Marsden NHS
Foundation Trust
6.5.2 Smoothing techniques from the University of Virginia
6.5.3 Smoothing technique from the Memorial Sloan Kettering
Cancer Institute
6.5.4 Smoothing technique from the University of Tubingen
6.5.5 Smoothing technique in the Nucletron PLATO TPS
6.5.6 Smoothing technique at the Thomas Jefferson University
6.5.7 Smoothing techniques at University of Maryland
6.5.8 Smoothing techniques at University of California, San
Francisco
6.5.9 Smoothing techniques at Sichuan University, China
6.5.10 Summary on smoothing
6.6 Incorporating MLC equipment constraints in inverse planning
6.7 Beam direction optimization
6.8 Monte Carlo dose calculation
6.8.1 The debate over the usefulness of Monte Carlo
dose-calculation techniques
6.8.2 Determination of photon spectrum and phase space data
for Monte Carlo calculations
6.8.3 Comparison of Monte Carlo and pencil-beam calculations
6.8.4 MCDOSE
6.8.5 Speeding up Monte Carlo dose calculations
6.8.6 Monte Carlo calculation accuracy and error
6.8.7 Application to the dMLC technique
6.8.8 Monte Carlo calculations in tomotherapy
6.8.9 Monte Carlo calculations of IMAT
6.8.10 Other reports on Monte Carlo dosimetry
6.9 Energy in IMRT
6.10 Measuring and accounting for patient/tumour movement
6.10.1 General review
6.10.2 Some observations of the effects of movement
6.10.3 Optical imaging for movement correction
6.10.3.1 Breast movement
6.10.4 X-ray imaging of position
6.10.4.1 Intrafraction and interfraction prostate
movement measurements
280
284
286
286
287
288
288
288
290
290
292
292
292
293
293
294
295
296
305
305
307
307
310
312
313
313
315
315
316
317
319
319
320
321
323
323
323
Contents
6.10.4.2 Intrafraction and interfraction lung movement
measurements
6.10.5 Ultrasound measurement of position
6.10.5.1 The NOMOS BAT
6.10.5.2 Other ultrasound systems developed
6.10.6 Magnetic monitoring of position
6.10.7 Gating
6.10.7.1 Gating based on optical measurements of
surface markers
6.10.7.2 Gating based on x-ray fluoroscopic
measurements
6.10.7.3 Imaging and therapy gated by respiration
monitor
6.10.7.4 Measurements using oscillating phantoms
6.10.7.5 Evidence against the need for gating
6.10.8 Robotic feedback
6.10.9 Held-breath self-gating
6.10.10 Intervention for immobilization
6.10.11 Active breathing control
6.10.12 Calculating the effect of tissue movement
6.10.12.1 Incorporating movement knowledge into the
inverse planning itself
6.10.12.2 Use of multiple CT datasets and adaptive IMRT
6.10.12.3 Modelling the effect of intrafraction movement
6.10.12.4 Modelling set-up inaccuracy
6.10.12.5 Modelling the movement of OARs
6.11 Megavoltage CT (MVCT) and kilovoltage CT (kVCT) for
position verification
6.11.1 MVCT
6.11.2 Flat-panel imaging for kVCT
6.12 MRI and IMRT simultaneously
6.13 IMRT using mixed photons and electrons
xiii
327
329
331
335
337
337
337
339
341
343
345
346
347
348
348
351
351
352
359
367
368
368
368
372
374
376
Epilogue
379
References
382
Preface and acknowledgments
Intensity-modulated radiation therapy (IMRT), virtually non-existent 20 years

ago and still embryonic 15 years ago is now an established clinical tool. Its
development, once the preserve of physicists and engineers, whilst incomplete,
is now at a stage when application specialists, oncologists, radiotherapists
and radiographers can introduce IMRT to the clinic. Meanwhile physicists
and engineers both continue research and development and support clinical
implementation. Feedback from clinical use in turn influences R and D. Clinical
R and D is the new IMRT specialty and randomized clinical trials of IMRT are
still only a handful.
Previous book reviews of IMRT (from me and from others) have largely
concentrated on its physics. This volume continues to review the developing
physics but also covers the preliminary clinical implementation and the interplay
between the two. The book can be read standalone and the chapters can be read
in any order. This fourth book in a series is, however, better read, and should be
more use, alongside the three complementary volumes. Only a limited repetition
of detailed physics is included. Consequently this volume is less mathematical
than its predecessors.
Multi-author books can draw on greater talent; the downside is that they are
also often disjointed, overlapped and sometimes impossible to create because of
so many competing priorities for authors time. Single-author books ought to
have a consistent style and cohesion but inevitably reflect the particular skills,
special interests, but also limitations, of the author. In some areas of this book,
I have reviewed clinical detail in which I relied on trusting my primary sources
rather than my internal quality control as a physicist. There is therefore a greater
weight of debateable opinion than objectively established fact than in the previous
volumes; but I hope we would welcome that. If there were ever to be a fifth
volume it should be a work jointly with a clinician. The references here are
timestamped with an end date of June 2004.
I am deeply grateful to innumerable colleagues and friends both at the Royal
Marsden NHS Foundation Trust and the Institute of Cancer Research and also in
hospitals and research labs throughout the world. Their work essentially creates
this review and they will recognize their contributions to the field herein.
I should like to thank John Navas, Simon Laurenson, and others at IOPP
(and their reviewers) who have put their faith in my efforts. I am very
grateful to Marion Barrell for much help with organizing figures, checking
references and seeking permissions to publish. I also thank Marion and our other
departmental secretaries (Rosemary Atkins, Sylvia Bouchere , Dana Roberts and
Lesley Brotherston) for typing so many dictated tapes during the past four years
of book writing. I thank all the authors and publishers who have allowed material
to be used here for illustration1. Contact was made with all copyright holders and
I apologise for those few cases where no response was received.
The material reviewed here represents the understanding and personal views
of the author offered in good faith. Formulae or statements should not be used in
any way concerned with the treatment of patients without checking on the part of
the user.
This book is dedicated to Linda.
Steve Webb
June 2004
1 Some figures only relate to their captions when viewed in colour. These are indicated in the figure
captions and colour figures can be found at http:/bookmark.iop.org/bookpge.htm?&isbn=0750310049
Chapter 1
Intensity-modulated radiation therapy
(IMRT): General statements and points of
debate
1.1 Observations on IMRT at the current time

The year 2000, representing the entry into a new millennium, has been highlighted
as a turning point in many contexts. Much of this is undoubtedly opportunistic
capitalizing on a round-figure date but it conveniently was the year in which IMRT
came of age and turned from a physicists and engineers interest into a rapidly
expanding clinical technique whose prosecution in the clinical arena required the
future collaboration of medical doctors, radiographers, engineers, physicists and
others. This in turn set the scene for shifting the focus of attention to the expected
clinical benefit. Meanwhile work continued to develop and refine the physics and
technology of IMRT.
Morita (2000) delivered an address to introduce an International Takahashi
conference that appropriately reminds us of this important scientist in the history
of conformal radiotherapy (CFRT) and IMRT. Shinji Takahashi was one of the
pioneers of radiological tomography (having invented various forms of axialtransverse tomography in the 1950s) and also of what he called conformational
radiotherapy (a technique invented in 1960). Takahashi linked the definition of
target volumes using axial-transverse tomography to his forms of conformational
radiotherapy. In the 1960s, conformational radiotherapy was achieved in Japan at
just a few centres using rotating multileaf collimators (MLCs) (with big leaf size)
and rotating protectors whilst the patient simultaneously corotated (figure 1.1).
These techniques were continually developed in Japan throughout the 1980s
and largely in the 1990s have been overtaken by inverse treatment-planning
techniques based on computed tomography and with the development of more
sophisticated IMRT. Morita (2000) summarized the chain of tasks to be optimized
to maximize the effectiveness of radiation therapy. Some specific cases were
given to demonstrate the improved performance of CFRT. Whilst not providing
Intensity-modulated radiation therapy (IMRT)
Figure 1.1. The principle of conformational radiotherapy as exemplified in the historical

work by Takahashi. In this figure the patient section is shown rotating on the right. A
MLC with large leaves, is shown at A2 defining a beams-eye view of the target. The
shape of the aperture is varied as the beams-eye view of the target changes. At A1
on the left, a radiation-attenuating protector is also shown changing its disposition with
respect to the collimator and the body as a function of angle. This is a form of binary
(onoff ) modulation and, whilst it achieves some conformality, is not as successful as
modern IMRT. (Reprinted from Morita (2000) with copyright permission from Elsevier.)
for high-spatial-resolution fluence variations, these Japanese developments are

widely cited as early pioneering attempts to improve the physical basis of
radiation therapy and should be regarded as part of the history of IMRT. Elsewhere
in the world most radiotherapy used simple fields with or without shaping with
blocks. Conformal therapy with automatic field shaping did not start until the mid
1980s and IMRT came a decade later (figure 1.2).
In the summer of 2000, I completed a summary of IMRT development to
that time (Webb 2000d) and this book begins where that one left off, reviewing
the work so far this century and looking to the future. That summary of IMRT
(Webb 2000d) was completed three weeks after the 13th International Conference
on the use of Computers in Radiotherapy (ICCR) in Heidelberg and, in the
appendix, I wrote the last line to the book: Where will IMRT be at the time
of the 14th ICCR (Seoul, South Korea)?. This volume was completed a few
weeks after this 14th ICCR and summarizes the answer to that question. There
have been several overviews of the field in the last four years. I provided a
response to an invitation at the turn of the millennia to crystal-ball gaze the
future of the physics of radiotherapy (Webb 2002e). In Seoul, Ha (2004) also
gave a set of predictions for what is important in the future of radiotherapy.
Observations on IMRT at the current time
(a)
Traditional radiotherapy with 4

orthogonal fields and no blocking
(b)
rotation therapy
with constant-area
field
IMRT
(c)
Figure 1.2. This figure encapsulates the key differences between the major forms of photon
radiotherapy. A representative PTV is shown as a solid comprising three cubic volumes.
Imagine also that this PTV is surrounded by normal tissue which it is desired to spare
from radiation. (a) shows how four uniform-intensity fields irradiating the volume normal
to its planar faces would create a rhomboidal (box)-shaped region of high dose. Clearly
volumes of normal tissue are irradiated. (It is accepted that the situation could be improved
by geometric field shaping). (b) shows the principle of rotation therapy in which a field of
uniform intensity and rectangular area is rotated through 2 around the PTV. A cylindrical
high-dose volume would result, again irradiating normal tissue unnecessarily. (c) When
fields are shaped geometrically using a MLC and also the intensity is varied across the
plane of each field, the high-dose volume can be sculpted to fit like a sheath around the
PTV sparing normal tissue. This is the goal of IMRT. The figure is of course simplistically
schematic. In clinical practice, the PTV may have a highly irregular shape. Whilst it can be
argued that careful adjustment of field parameters might make conventional radiotherapy
acceptable for this geometrical volume, for more contorted PTVs this would not be possible
without CFRT or IMRT. Also for IMRT the degree of conformality will depend on the
number and intensity structure of the fields as well as on the shape of the PTV and its
proximity to normal tissue.
The IMRT Cooperative Working Group generated a lengthy review of IMRT

in November 2001 (IMRTCWG 2001) which can be read as a tutorial on the
subject and the IMRT Subcommittee of the American Association of Physicists
in Medicine (AAPM) Radiation Therapy Committee has also produced a lengthy
report offering guidance on IMRT implementation (Ezzell et al 2003). A joint
document from the American Society for Therapeutic Radiology and Oncology
(ASTRO) and AAPM has overviewed the whole process of implementing clinical
IMRT (Galvin et al 2004). I have also provided a brief history of the developments
in improving the physical basis of radiotherapy including comment on IMRT
(Webb 2002d) and a longer history of IMRT (Webb 2003a). Table 1.1 provides
a summary of key historical landmarks in the development of IMRT. In 2003 the
AAPM held its annual Summer School entirely focusing on IMRT and resulting
in a very comprehensive monograph (Palta and Mackie 2003). In the winter
months of 2003/2004, the British Institute of Radiology published a seven-part
set of review articles on IMRT in the British Journal of Radiology (Webb 2003b,
Williams 2003b, McNair et al 2003b, James et al 2003, Guerrero Urbano and
Nutting 2004a, b, Beavis 2004b).
At the risk of being quite controversial, I begin by standing back and taking
a long view on IMRT and maybe on the whole subject of improving the physical
basis of radiotherapy. I hope I am forgiven for sharing this somewhat personal
view. It is offered as a reflection on the history of IMRT (Webb 2003a, b) and in
the spirit of seeking a positive way forward. Perhaps it can be read as one might
read a similar editorial to a journal packed with interesting articles on IMRT?
The principle of IMRT, that pencils of radiation emanating from different
directions and with different intensities can be joined together to create high
dose volumes with concavities in the outline and spatially-varying high dose,
has been linked to an interesting historical analogy from Birkhoff (1940). He
showed that, provided negative pencil blackness could be used, any picture could
be represented by the linear sum of straight pencil lines of different blackness
from different directions. Figure 1.3 shows his, now famous, cat picture created
this way. Of course, negative x-rays do not exist so the exact analogy breaks
down.
Towards the end of the 1980s, very few physicists had begun to consider
that the possibility to modulate the fluence of a beam would expand the options
to create a high-dose volume around the tumour whilst sparing organs at risk
(OARs). The notion of what today we call IMRT (which was incidentally not
called that at the time) is generally credited to the Swedish medical physicist
Anders Brahme. Some (maybe even he?) would argue that there were earlier
attempts to modulate fluence via the tracking cobalt unit, the gravity-driven
blocking mechanisms and other ways. However, these were very simple binary
onoff modulations of the primary fluence (ignoring leakage) and it is to Brahme
that the idea of inverse-planning for modulating fields is rightly credited. He
solved one of the only problems amenable to analytic inversion in 1982 and his
1988 paper is considered the seminal first work on inverse planning. Thomas
Table 1.1. The development of IMRTthe one-page history. IMRT is poised to make a
major clinical impact. How has the field reached the present position?
1895
1896
1950s
1959
1960s
1970s
1982
1984
1988
1988
1989
1990
1991
1992
1992
1993
1994
1994
1999
2002
2004
The x ray was discovered on 8th November in Germany by Rontgen.

Doctors understood the need to concentrate radiation at the target but had
means neither to do this nor even to know where the target was precisely.
Takahashi first discussed conformation therapy.
Invention and patenting of the first multileaf collimator.
Proimos developed gravity-oriented blocking and conformal field shaping.
The Royal Free Hospital built the tracking cobalt unit and MGH Boston did
similarly.
Brahme et al discussed inverse-planning for a fairly special case of rotational
symmetry.
First commercial MLCs appeared.
Brahme published first paper on algebraic inverse planning.
Kallman postulated dynamic therapy with moving jaws.
Webb developed simulated annealing for inverse planning. So did Mageras and
Mohan.
Bortfeld developed algebraic/iterative inverse-planning, the precursor of the
KONRAD treatment-planning system.
Principle of segmented-field therapy developed (Boyer/Webb).
Convery showed the dMLC technique was possible.
Mark Carol first showed the NOMOS MIMiC and associated PEACOCKPLAN
planning system (now CORVUS).
Tomotherapy (the Wisconsin machine) first described by Mackie.
Stein, Svensson and Spirou independently discovered the optimal dMLC
trajectory equations.
Bortfeld and Boyer conducted the first multiple-static-field (MSF) experiments.
First discussion of possible robotic IMRT.
Commercial tomotherapy began.
Large number of competing IMRT planning systems and systems for delivery.
IMRT is well established in several geographically distributed centres.
Bortfeld, Radhe Mohan and I were independently working out ways to create
modulated fluence in 1988 and by 1990 there were just a handful of papers on
IMRT; the techniques were entirely concerned with planning not delivery. I do
not think any of us imagined how fast and how greatly this research field would
develop. Certainly I did not for I am no prophet.
The development of IMRT has been astonishingly fast. By the mid-1990s
the essential features of all the main techniques for delivering IMRT had been
established (see reviews in Webb 1993, 1997). Certainly there were details
still to unravel, and important ones at that, but the key papers on the physics
of IMRT delivery had been written. All the developments, with the exception
of the NOMOS MIMiC, comprised one-off techniques developed in universities
Figure 1.3. The diagram in the paper by Birkhoff (1940) showing how a black-and-white
picture of a cat can be made up from a series of pencil lines of different thickness (density)
provided these are both positive (normal pencil) and negative (unphysical erasure). This
may be regarded as analogous to the principle of IMRT except for the fact that there are,
unfortunately, no negative x-rays. (Reprinted from Journal des Mathematiques Pures et
Appliqees 19 22136.)
or hospitals. There was nothing available from manufacturers (again with the
exception of the MIMiC).
By the turn of the millennia, all this had dramatically changed (see review
in Webb 2000d). The major accelerator manufacturers were each offering IMRT
delivery packages. Most of the treatment-planning system manufacturers were
also offering IMRT planning facilities, usually by inverse planning. Focus
began to shift from fundamental concerns about how to perform IMRT into the
embryonic stages of clinical delivery. The fundamental developments of IMRT
could be consigned to history, albeit very recent history (Webb 2003a).
All this is very satisfying because it shows that the transition from
just a few workers doing fundamental physics through to widespread clinical
implementation had been achieved in about 15 years which is a very short time
against the background of the implementation of other technologies.
The scale of research progress has created a difficult position with respect to
the literature. There are now literally thousands of papers and abstracts of work
on IMRT and these can be hard to assimilate. Many papers report marginally
different solutions to the same problems. One of the reasons for this and previous
books in the series has been to try to overview these, filter them and present them
to a new audience. This distillation is now, in my opinion, close to impossible.
Also there are hundreds of papers and abstracts showing how commercial
inverse planning has been coupled to commercial IMRT delivery and used
to initiate clinical IMRT. Again, this is very pleasing in that it indicates the
widespread adoption of the concepts. But many of these reports are remarkably
similar.
Is it too harsh to ask if there is a climate of over-reporting? The phrase
write-only journals has crept into the language! I am not old enough to know
whether the same phenomena arose when, for example, the first linacs began to be
used and others then copied the installations. But the slimness of back copies of
the major journals together with the fact there were less meetings and conferences
suggests that it did not.
There are also those who are orchestrating campaigns to challenge the
clinical significance of implementing IMRT (not oppose it but question the
possibly previously unquestioned speed of implementation and justification).
There have been several important contributions from Schulz whose main thesis is
that IMRT implementation is more due to manufacturer pressure, reimbursement
and copycat fashion than from the basis of genuine proof of clinical benefit (see
section 1.2). His perspective may be more American than European where phase3 trials are being organized. This was also mentioned by Beavis (2003b) when
describing implementation in an NHS clinic.
What is the way forward? Firstly I believe there are still many unsolved
problems with the fundamental physics of IMRT and it is right to continue to
investigate these. An example is the somewhat approximate nature of predicting
the dose using analytic expressions and/or an incomplete knowledge of the
behaviour of scatter dose with small-area segments. Another example is to
challenge the assumption that one can ignore tissue inhomogeneities at planning
and delivery (Papanikolaou et al 2003). Monte Carlo (MC) approaches will
address that. Next, we still have less than adequate methods to check and measure
absolute dose and to find monitor unit (MU) calibrations. There is the whole issue
of planning for the moving patient rather than on some single imaging event (as
if the patient remained frozen as imaged). One can identify that there could be a
requirement to simplify the whole procedure of IMRT delivery by using purposebuilt equipment rather than the MLC. Possibly robotic delivery could increase
treatment flexibility. IMRT using cobalt units has hardly been addressed yet. This
might open up IMRT for departments or countries without an adequate supply of

linacs.
Regarding clinical studies, there are many who simply report the
implementation of a technique on the basis that the predicted dose distributions
will be better. This is admirable but somehow incomplete. What we really want
to know is whether the implementation will make a difference and, even if it
does, whether this difference is significant in terms of benefit to the individual
and costbenefit to society. According to Sharpe (2003) Confidence comes from
evidence. For this reason, I believe the clinical papers that merit most attention
are those that are accompanied by documented evidence of either reduced side
effects or improved probability of cure. Even better will be the reports of phase-3
clinical trials and, at the time of writing, there are not many of these (Jayaraman
2003). However, in chapter 5, an effort has been made to clearly identify the
few papers reporting genuine clinical benefit and distinguish them from paper
planning exercises or clinical implementations with no reported findings.
Perhaps the really long view comes from the unadorned observation that
radiotherapy is only a matter of placing some required dose at the right place
and avoiding doing so at the wrong place. It is 108 years since people started
to try to do this in 1896, the year after the discovery of the x-ray. No-one yet
believes it is done 100% properly. Literally thousands of workers have contributed
to aspects of the problem. So it would be unthinkable that any of us right now
should regard the matter as not worthy of further study.
IMRT is just one wayand a very fine wayto relocate dose spatially. This
hardly equates to having cracked all the problems of radiotherapy. It is important
to focus on unanswered questions and novel opportunities.
The philosophy of IMRT and its ultimate dream scenario has been discussed
(Webb 2001c), in particular noting that the goal has changed from seeking the
ultimate conformality at the price of great complexity to achieving practical
clinical gain. Current constraints are both technological, humanitarian and
financial. A two-track approach is recommended in which there will be
a continued development of fundamental concepts in tandem with clinically
extending known techniques. Geographical widening of IMRT availability is
an issue. Possibly the MIMiC-type IMRT may be overtaken by dynamic MLC
techniques.
Bortfeld (2001) has reviewed physical and technological aspects of IMRT.
Current issues at the cutting edge include: improvement of the accuracy of dosecalculation algorithms, Monte Carlo planning, reduction of the number of field
segments and improved efficiency, use of MLCs with smaller-than-standard leaf
widths, use of MLCs with small shifts of the isocentre or rotation and accounting
for organ motion and inverse planning and delivery.
Suit (2002) in his L H Gray Lecture argued forcibly for the view that, in the
next two to three decades, protons will entirely replace photons as the treatment
modality of choice. This is predicated on the argument that dose distributions
are incontrovertibly superior. There is no clinical debate. The only concern
is financial. Proton facilities cost more. However, it may be argued they are
in service longer. The Harvard Cyclotron operated continuously for 52 years.
One accelerator can serve many gantries. All adjunct costs are the same as for
photons. The improvement in disease definition through functional imaging is
Suits second prediction. A third major area of development will be that of
techniques to compensate for internal organ movement. He highlights the history
of doubting Thomases. Almost all developments in improving the physical basis
of irradiation therapy have had their critics. He urges us to follow the maxim of
the pioneering physicist Lord Rutherford, known as the old Croc, because of his
interest only in the future. Crocodiles cannot look backwards. He also writes:
The history of medicine has repeatedly demonstrated that the perceived efficacy,
and not the cost, primarily determine the fate of new technologies.
Williams (2002a) has discussed how IMRT has moved from the sole preserve
of the research community to a wider clinical application. His view is that the
choice between dynamic and step-and-shoot delivery and the choice between
forward and inverse planning has less to do with the overall outcomes as measured
by the three-dimensional (3D) dose distribution than it has to do with local choices
and local availability of equipment. Williams (2002b) has identified image-guided
radiotherapy (IGRT) as perhaps the final missing link in our ability to deliver
high-dose volumes to planning target volumes (PTVs) which are built with very
small margins. One might comment that understanding the functional status of
the target, sub-targets and OARs is also a key element. Accommodating the live
moving patient is also still a problem to be solved.
Walker (2001) outlines the steps in performing and issues surrounding, the
development of clinical CFRT and the role of IMRT. Balycyki et al (2001) discuss
the issues underpinning the selection of new linear accelerators. One of the
criteria highlighted is the need to support IMRT delivery. They note that this
is difficult to arrange when the field is so rapidly moving. Cost should not
be the main preoccupation. A study by Bate et al (2002) showed that when
implementing IMRT of the prostate there was an increased workload at the level
of simulation and planning but that, except in the first four days, the treatment
time did not increase.
Most of us predicate the view that, as technology becomes more complex,
yet more staff will be needed to run it, to check its accuracy, to ensure no errors
and so on. A recent Medical Physics Point and Counterpoint between Starkshall
and Sherouse has debated this (Starkshall et al 2001). Sherouse believes in this.
He believes the role of the medical physicist is to be much wider than some
quality control engineer. He quotes Arthur C Clarke that sufficiently advanced
technology is indistinguishable from magic and that this requires the qualified
magician to be much more than a sorcerers apprentice. The difference would be
revealed in a radiation crisis of the kind of which there have been several noted
examples. He feels the aspects of caring for machines have been overemphasized
compared with caring for people.
10
Conversely, Starkshall believes new biological warfare on cancer will make

radiation therapy redundant. He believes modern machinery can essentially check
itself and requires little more than spot checks. He believes in treat-by-wire
like fly-by-wire. He does not believe medical physicists would avert disaster and
should certainly not sit around waiting for the emergency which never arises.
There is much to debate (see also Webb [2000d] on this same subject).
IMRT is coming of age. Around the world, courses are now regularly
held. Among these are the (Siemens) MRC-Heidelberg IMRT Schools, the Royal
Marsden School, the School at The Netherlands Cancer Institute, the School at
the Mallinkrodt Institute, St Louis, the School at Stanford University and others.
Other equipment manufacturers also run IMRT Schools. For example, a report on
the first Elekta-sponsored course (March 2000) on IMRT appeared in Wavelength
(2000d) in July 2000.
In summary, this introduction has highlighted:
(i)
(ii)
(iii)
(iv)
the early beginnings of IMRT;

the very rapid development in which history seems almost a misnomer;
the burgeoning literature and suspicion of overpublication;
the focus on image-guided radiotherapy as the largest unsolved problem and
imaging for plannning and imaging for accounting for the moving patient;
(iv) the switch from technological development to the proliferation of clinical
IMRT;
(v) the burden of the cost of new technology, unlikely to diminish in the short
term;
(vii) the growth of IMRT Schools;
(viii) the growth of an IMRT backlash.
1.2 Criticism of the philosophy of IMRT

In this section, we spend some time with the issues of IMRT philosophy and
specially concentrate on those writings which have generated criticism of the
clinical implementation of IMRT.
In November 2000, the journal Medical Physics pitted Sarah Donaldson
against Art Boyer on the subject of the wisdom of introducing IMRT widely
(Donaldson et al 2000). Donaldson argued:
(1)
(2)
(3)
(4)
(5)
that there are little data to support general non-investigative use;

that there are too many inaccuracies associated with patient positioning;
that organ motion remains beyond physician and physicist control;
that quality assurance procedures for IMRT are in their infancy;
that IMRT requires unusually and unacceptably long delivery times that are
disruptive in a busy clinical department;
(6) that IMRT involves excessive leakage radiation;
(7) that fusion of CT and magnetic resonance imaging is imprecise for IMRT;
Criticism of the philosophy of IMRT
11
(8) that IMRT start-up and maintenance costs are too expensive and
(9) that there is a costly learning curve for IMRT.
She argues that IMRT should remain an institutional investigative tool only.
Against the proposition, Boyer claims that the lack of biological data has
no special role in IMRT any more than in any other form of radiation therapy.
Todays therapy techniques have been developed against a background of lack of
such knowledge and the same should happen for IMRT in the future. Secondly,
he refers to the Catch-22 whereby, unless there are extensive IMRT facilities,
the needed experience to enable clinicians to use 3D dose information and the
additional control over the modulation will not be gained. He also points out that,
at least in the United States, there is a wide variety of institutional techniques with
a range from the very simple to the very sophisticated.
Donaldson then counters that IMRT adversely impacts the current working
practices of the radiotherapy department. She concludes that the technique should
only be applied in a research setting. In practice, many clinics have found
that IMRT can be performed without increased burden once the initial stages of
implementation have been worked through.
Conversely, Boyer points out an important issue, namely the preferred scope
and speed of implementation of IMRT. If IMRT is introduced too early, this could
precipitate experiences that will condemn the process. However, implementing
IMRT too slowly will unduly postpone the realization of the advantages of the
technique. Boyer argues that the wider the implementation of IMRT is, the more
likely we are to answer the questions that concern Donaldson sooner rather than
later. Similar arguments were made by Allen Lichter for the introduction of
conformal 3D radiotherapy in the early 1990s.
This topical debate underlines concerns about the clinical implementation
of IMRT that were heard hardly at all three to four years ago. Despite the
fact that IMRT was being developed, little except positive comment had come
from doctors and clinical physicists keen to take advantage of the approach.
Most, fortunately in my opinion, still express this keenness and doubters are in
the minority. This section highlights them to encourage the reader to enter the
debate. The logic of the pro-IMRT-ers is not over-emphasized here because it is
represented by the scientific contributions in the rest of the book. The remainder
of this section may thus appear quite negative to IMRT.
Paliwal et al (2004) enter the debate on the wisdom of heavy use of IMRT.
One protagonist in this Medical Physics Point and Counterpoint argues that its use
(in the USA) has more to do with gaining profits from reimbursements and less
to do with demonstrable improvement in tumour control. The counter protagonist
argues that using IMRT for 20% of patients is justified and that IMRT properly
continues the quest for improved precision.
Boyer (2001) has emphasized that implementing IMRT requires an
assessment of primary and secondary barrier radiation safety as well as
consideration of the risk of inducing a second carcinoma to the patient. Magnetic
12
resonance images should be incorporated into the treatment-planning process

and all staff should understand the, at least temporary, increase in work burden
as a result of implementing IMRT. Boyer (2001) argues that the increased
reimbursement for IMRT should be used to provide the necessary additional
human resources. Good estimates of these costs have been provided by Gillin
(2003).
Potters et al (2003b) question the definition of IMRT in the context of an
editorial commenting on a specific paper. For example, they refuse to accept that
breast irradiation using several static-MLC fields constitutes IMRT of the breast
in the context of the definitions of process of care required for US Healthcare
reimbursement. Vicini et al (2004) argue that concentrating attention on the
terminology has detracted from the importance of improving breast IMRT in
clinical practice. The argument is uniquely relevant to a healthcare system based
on insurance reimbursement.
Purdy and Michalski (2001) have provided an editorial discussion on
whether the clinical evidence for CFRT and IMRT supports the enthusiasm for the
technology. They point to two important clinical trials. One, at the MD Anderson
Hospital, compared delivery of 70 Gy or 78 Gy to the prostate with CFRT in a
phase-3 randomized trial. The outcome showed increased freedom from disease
failure and biochemical failure at the higher dose (Pollack et al 2000, 2002a,
Pollack and Price 2003). A second study (Ryu et al 2001) showed favourable
lower toxicity for increased dose. (Further hard clinical evidence may be found
in chapter 5.)
Purdy and Michalski (2001) also emphasized that 3D radiotherapy does not
mean using non-coplanar beams. It does mean taking more care over the 3D
specification of the gross tumour volume (GTV), clinical tumour volume (CTV)
and PTV. They feel that medical doctors should embrace this technology, become
better educated and start to use it even before there is overwhelming evidence of
clinical efficacy. Their view is that there is no other path to improving the physical
basis of radiation therapy.
Levitt and Khan (2002) are, conversely, sceptical of the rush to judgement in
favour of CFRT and IMRT. They point to difficulties in determining the volumes,
problems with assessing the outcomes of radiotherapy and limitations with the
few extant trials.
The several writings from Schulz address the same concerns about the speed
of IMRT clinical adoption. A Medical Physics Point and Counterpoint (Schulz
et al 2001) has been given to the controversial arguments from Schulz that the
pursuit of IMRT is falsely based and possibly a waste of time. Schulz argues
that IMRT is highly profitable for manufacturers and good for employment of
physicists but will not significantly reduce overall cancer mortality. He provides
detailed statistics showing that IMRT at best can reduce the number of deaths
by about 4900 in the United States, for which each centre would need an IMRT
system, each of which would at best extend the lives of five patients per year. He
concludes that this capital investment and operational expense cannot be justified.
13
Arguing against this proposition, Deye makes the point that, first of all, all new
technologies have always had their critics and the arguments raised are not unique
to IMRT. He also argues that outcome data cannot be used to predict which
new technologies would be beneficial. He also argues that if IMRT is seen to
be being introduced too quickly then this would imply that physicists are not
spending enough time on the new developments rather than too much. In short,
his argument is that it is the duty of scientists to get the science right and to
let those who deal with the big picture put things into prospective. Deye also
remarked that there is clear evidence for reduced rectal toxicity in treating prostate
cancer from the work at Memorial Sloan Kettering Cancer Center (see chapter 5).
Schulz sticks to his guns and argues that it is competition between hospitals for
patients that plays a far larger role in purchase decisions than do expectations
of improved clinical outcome and that, in short, IMRT is market driven. This
argument (Schulz et al 2001) is almost an exact repeat of one a few years earlier
from the same author (see review in Webb 2000d).
Then, just a few weeks later in the literature (Schulz and Kagan 2002a),
the same statements criticizing the overfast adoption of IMRT receive further
printspace along with a general criticism of the outcomes of a workshop designed
to find the most effective areas of medical physics research. The comment
extends to criticize the programmes of work on image-guided radiotherapy and
the Cyberknife. The rebuttal is less than firm (Cumberlin et al 2002) but cautious.
Schulz and Kagan (2002b) have another attack at the enthusiasm for IMRT, this
time with a reasoned discussion of the clinical issues in each tumour site. The final
paragraphs intended as tutorial for radiation physicists are particularly barbed.
However, the paper does provide a systematic comment on an organ-by-organ
basis.
Schulz and Kagan (2003a), now becoming familiar critics of IMRT,
comment on IMRT in cancer of the prostate. They state that, given the margins
applied to prostate to create a PTV, the rectum must be significantly inside the
PTV at least on its anterior side and therefore they find it hard to understand
why sculpting the dose around the PTV will lead to an improvement in rectal
complications. They argue that the rectum will wander in and out of the high-dose
region due to normal physiological function. They even think that the reduced
rectal toxicity might result as much from organ motion as from the precisely
positioned dose distribution delivered by IMRT. In response to this, Zelefsky
and Leibel (2003) re-present the evidence that IMRT of the prostate leads to
decreased rates of Grade-2 rectal bleeding and they claim that this is explained by
the reduced circumferential volume of normal rectum exposed to high radiation
dose levels. They agree that prostates have always moved and will continue to do
so and that this movement is not the cause of reduced rectal complications (see
also chapter 5).
Kagan and Schulz (2003) and Pollack et al (2002b) further debate whether
a change in biochemical failure is related to cause-specific death in prostate
cancer. Pollack et al (2002a) found a change in post-treatment PSA levels when
14
escalating CFRT from 70 to 78 Gy at Fox Chase Cancer Center (FCCC). Keall

and Williamson (2003) have defended the development of IMRT technology in
the light of further criticism from Schulz and Kagan (2003b) who claim that the
ultimate radiation therapy (the infinitron) would still be inferior to surgery.
Fraass (2001) has commented that IMRT and inverse planning have been
accepted somewhat quickly with few studies which quantitatively compare the
benefits of this new technology with older CFRT techniques. Somewhat more
positively he presented a number of types of comparison experiment which could
be used to address most of the limitations that apply to making comparisons
between unlike treatment modalities.
Glatstein (2003a) has written an interesting piece pointing out that the late
20th century has been a unique time for hyping technology (and drugs) directly
to the public with almost no proof of their efficacy available. He cautions against
this and is concerned that the dollar drive may be behind it. Many patients are
not able to assess the information they receive. The new techniques are billed
as routine rather than research because only the former can be charged for in
the USA. In response to Glatstein (2003a), Amols (2003) reminds him and us
that whilst it may be convenient to blame manufacturers, administrators, the
internet, the mass media and patients themselves for the hype and the growth of
IMRT, it is the clinician who holds the key to the IMRT suite and thus controls
the volume of use of this modality. Glatstein (2003b) reports that IMRT will
not be subject to controlled clinical trials because to do so would challenge the
notion that IMRT is clinically reimburseable and instead is research. Also he
points out we live in an era when direct appeal is made to the sick patient to make
treatment choices they cannot comprehend. Some American insurance providers
are refusing to pay for IMRT until the results of randomized controlled trials
are known (Potters et al 2003b), a peculiarly perverse situation given both the
paucity of such data at present and the view expressed previously that, in some
circumstances, it may be unethical to conduct such trials. This is a peculiarly
American perspective because clinical trials are planned in Europe. Nutting and
Harrington (2004) and Dearnaley (2004) have joined this correspondence to point
out that, in the UK, precisely the reverse economic conditions prevail. Unless a
technology can be proven worthwhile, the National Health Service (NHS) will
not fund it. They look forward therefore to the results of such trials and sharing
this information with colleagues in the USA to address this issue. Amols (2004),
however, emphasizes that even if (when?) proved valuable, IMRT issues of cost
effectiveness will still remain. Also, not surprisingly, he points out it is convenient
for the less well-funded nation not to have gone the way of the USA.
Halperin (2000) has also questioned the uncontrolled growth of radiation
technology. His view is that healthcare costs will grow to 17% of GDP in the
USA by 2010 just as the baby boomers retire. He is wary of unscrupulous
salesmen who (sic) crank up sales or threaten to cut off the oxygen of patient
custom. He questions whether overcomplex treatment plans are being prepared
to recoup more costs when simpler ones would suffice. I quote The distinctions
15
between evidence-based medicine and if-its-reimbursable-well-do-it medicine

have been lost. Radiation oncology, as a specialty, is guilty of encouraging
profligate spending, we will engender a backlash, and well deserve it. He does
comment that CFRT is a testable hypothesis. Indeed many groups are testing
it. Hutchison and Halperin (2002) have demonstrated how decisions regarding
radiation oncology purchases can be influenced by market-driven advertising.
Kavanagh (2003) wrote a provocative letter The emperors new isodose
curves which, if I have understood correctly, is advancing the view that IMRT
physicists are pursuing some clinically irrelevant goals. He likens the practice of
IMRT to the selling of art and postulates that the art dealer fares better than the
art critic.
There has been considerable debate following the original paper by Levitt
and Khan (2001) asking if the evidence supports the enthusiasm for CFRT of the
prostate and dose escalation. The editorial accompanying the article felt their
paper was too anti- technology (Purdy and Michalski 2001). Levitt and Khan
(2002) wrote a letter reiterating that their concerns lay with the inappropriate and
untrained use of the technology rather than with the technology itself. Also they
believed there is no clinical evidence for irradiating the prostate beyond 74 Gy and
that indeed such irradiation may lead to long term radiation damage of normal
tissues that will not be seen for many years. Purdy and Michalski (2002) reemphasize their beliefs in the need for better training and randomized trials.
In Colorado Springs, in June 2003, the AAPM held its annual Summer
School on the theme of IMRT. At that meeting there was a thorough exposition
of the state of the art (Palta and Mackie 2003). Whilst the physics is
globally universal, the emphasis was perhaps understandably from an American
perspective so far as implementation is concerned. In the discussions some very
quotable soundbites were heard including the following:
(i) Is IMRT tailored treatment or Emperors New Clothes?,
(ii) Is an IMRT plan really optimal or an optimal illusion?,
(iii) Remember there is a con in concave (Langer 2003).
These comments were not necessarily expressing the views of the speaker but
said to provoke discussion. With regard to escalating costs in IMRT technology,
Battista and Baumann (2003) defined some new principles in the ALARA
mould. Machines would be developed as long as it is technically achievable
(ALATA) but to operate the ALATA principle you need ALATA (a lot of)
money. Dose distributions do not necessarily have to be optimum but should
be as conformal as reasonably achievable (ACARA). Battista and Baumann
(2003) reminded us of a quote from Harold Johns, one of the radiotherapy physics
pioneers, that is as true today as when he said it: If you cant see it you cant hit
it and if you cant hit it you cant cure it. This was an exceptional School.
In concluding this chapter, which has been more about the politics of, and
belief systems in, healthcare, we may note that the first few years of the new
millennium have witnessed some published scepticism of the usefulness of IMRT.
16
Table 1.2. Arguments for and against the clinical implementation of IMRT.
FOR
AGAINST
(As expressed but not necessarily true)
IMRT creates concave high-dose

distributions protecting organs-at-risk
and enabling dose escalation to target.
The value of clinical implementation

is out of all proportion to the current
clinical evidence for efficacy.
IMRT builds on experience with 3D

CFRT.
Early clinical evidence for IMRT

supports the view that better dose
distributions lead to better clinical
outcome (e.g. reduced xerostomia,
reduced rectal toxicity, reduced dryeye syndrome, reduced paediatric
complications).
It is implemented: (i) because it is

there. (ii) because of manufacturer
pressure. (iii) because of $ reimbursement.
There are too many difficulties:

small physical segments, small MUs
per segment, moving parts, hard-toverify absolute dosimetry.
The moving patient destroys the

concept.
Workloads increase with IMRT.
Clinical implementation builds on

years of physics development which
has improved understanding of planning and delivery.
IMRT is hard to QA and verify.
It is thoroughly validated by professional groups and working parties.
Delivery times are long; possible

secondary cancers?
It is now relatively straightforward

and widely commercially available.
Mazes need redesigning.
IMRT is cost ineffective.
The public are being asked to choose

between treatments they cannot understand due to direct hype.
IMRT divides society.

mans technology.
Widespread implementation will allow randomized clinical trials which

will generate more evidence for benefit and eventually economies of scale.
Movement can be accommodated
using IGRT.
QA and verification tools exist.
It is a rich
17
This has ranged from what, at least to my eyes, looks like outright rejection of
IMRT as a useful clinical tool through to more balanced questioning of its role.
It is possible to be sympathetic to some of the underlying concerns about lack of
evidence. However, a climate of reasonable expectation of benefit is often all one
can put upfront and ultimately the proof of the worth of IMRT will lie in the
outcome of randomized phase-3 clinical trials. Without the resources to conduct
such trials, the evidence can never come and, on this basis, the proliferation of
technical installation and clinical implementation is justified. Many would say
this same philosophy of implementation without hard evidence was applied to
most of the significant developments in medical physics, e.g. the installation of
linacs instead of kV radiation, the development and use of treatment planning, the
coupling of 3D medical image data to treatment planning etc.
Table 1.2 summarizes some of the arguments for and against the clinical
implementation of IMRT. The table is not meant to express incontrovertible truths
but to offer arguments some have put. Readers can use this as the basis for their
own thought and discussion.
The bottom line, however, is that IMRT has an unstoppable momentum.
Chapter 2
Developments in rotation IMRT and
tomotherapy
The technologies of delivering IMRT in slices are known as rotation IMRT

or tomotherapy. Purists may argue about the precise terminology but serial
tomotherapy (MIMiC-based) and spiral tomotherapy (University of Wisconsin
machine-based) are related and usually reviewed together. The history of what
interaction there might be considered to be was best reviewed by Mackie et al
(2003b). The concept of the binary modulator for slit-field radiation was first
proposed by Swerdloff in Mackies group in 1988. This led to the patent for the
multivane intensity-modulating collimator (MIMiC) being held by the University
of Wisconsin Alumni Research Foundation (WARF) and licenced to NOMOS
who first made it a practical reality in 1992. The idea of spiral tomotherapy did
not come until about 1992 when spiral CT became possible and Mackie saw the
opportunity to reuse a gantry for this purpose. By then the NOMOS MIMiC was
a reality (NOMOS is now part of North American Scientific). It was another 10
years before the spiral tomotherapy unit, a much grander engineering venture,
became commercially available.
2.1 NOMOS MIMiC tomotherapy

2.1.1 Technology history
The NOMOS MIMiC technique for IMRT with its associated planning system,
originally known as PEACOCKPLAN, now as CORVUS, was introduced 12
years ago in 1992. It has been stated that its inventor Carol had the concept
as early as 1976 but did not take it forward at that time (Curran 2003). From
the very outset, it was a commercially available device. When announced in
the autumn of 1992 (ASTRO in Calgary and at a 3D radiotherapy meeting at
the World Health Organization [WHO] in Geneva), it took the radiation physics
community by some surprise because it was both unexpected and an unusual
NOMOS MIMiC tomotherapy
19
Figure 2.1. The serial tomotherapy approach to IMRT. This form of IMRT uses a
temporally modulated mini MLC system such as the MIMiC (NOMOS Corporation)
shown here mounted to a conventional low-energy megavoltage medical linear accelerator.
Treatment to a narrow slice of the patient is delivered by arc rotation. The complete
treatment is accomplished by serial delivery to adjoining axial slices. The right-hand part
of the figure shows a patients-eye view up into the MIMiC. (Reprinted from IMRTCWG
(2001) with copyright permission from Elsevier.)
concept. The MIMiC (figure 2.1) was designed to be attached to any linac. It
irradiated two narrow slices of the patient at any time by rotating a slit collimator
through a series of gantry angles. In its simplest form, the modulation switched
effectively every 5 (figure 2.2). From the patients-eye view, the machine looked
like a set of chattering teeth. The modulation was provided by varying the dwell
time of the attenuating elements in the slit. The variation was controlled by
an onboard rotating computer holding a floppy disk created from the planning
system. The machine monitored the gantry angle independently and the sequence
was recorded for verification. To irradiate a larger tumour, the patient was
sequentially longitudinally traversed through the line of sight of the MIMiC.
Until the first clinical IMRT made with an MLC in the late 1990s, use of the
MIMiC was the only way to deliver IMRT clinically and the NOMOS Corporation
could rightly style themselves as the Intensity Modulation Company. As MLCbased IMRT became more common, NOMOS diversified into planning for MLCbased IMRT. No-one really knows the exact number of IMRT treatments but,
at the turn of the Millennium, it could be confidently stated that more IMRT
treatments had been delivered by the MIMiC than by the MLC. One suspects that
that situation may have now changed with the rapid proliferation of MLC-based
IMRT. Detailed descriptions and photographs of the technology and developments
from the NOMOS Corporation appear in earlier books (Webb 1997, 2000d) and
20
50
MIMiC set
to 10
patterns,
changing
every 0.50
PTV
Figure 2.2. Schematic diagram showing the principle of operation of slice-based (MIMiC)
tomotherapy. The irradiation of just one slice is shown and so also just one bank of the
two-bank MIMiC is shown. The planning system works out the required 1D modulations
at each 5 of rotation around the patient. A possible modulation is shown at the right
of the figure. In order to realize this modulation in practice, the opening and shutting
pattern of the vanes of the MIMiC is adjusted every 0.5 as shown in the ten schematic
diagrams of the MIMiC (a dark bixel is meant to symbolize a closed bixel and a light
bixel symbolizes an open bixel). In view of the long-lever effect of radiotherapy, the
PTV sees the modulation as if it were at fixed 5 increments. After a rotation around the
patient, each of the two slices irradiated simultaneously by the two banks of the MIMiC are
appropriately irradiated. To irradiate the whole PTV, the patient is then shunted accurately
to the next longitudinal position. The drawing is not to scale because the depth of the
MIMiC vanes in the direction of the radiation is not properly represented here.
21
will not be repeated here. In the next subsections, we concentrate on issues

discussed since 2000.
2.1.2 Matchline concerns and solutions
From the outset there was concern about the reliability of matching the slices and
the Company provided the CRANE to ensure this. Even so, the technique may
have been safest when applied to the head-and-neck where patient immobilization
could be ensured. Dogan et al (2000a) addressed the match-line issue in serialslice tomotherapy using the NOMOS MIMiC. They suggested that each treatment
should be divided into two, one with the original isocentre determined by
CORVUS and the other with an isocentre shifted by 5 mm. Each treatment plan
would be delivered on alternate days summing to the total treatment dose. They
made measurements to show that the dose inhomogeneity in the PTV decreased
from about 25% to about 12% on doing this. The trick in the technique lay in
forcing CORVUS to offset the isocentre longitudinally by artificially adding a
length to the PTV in the longitudinal direction. The new technique was shown
to be more robust with respect to errors in longitudinal movement caused by the
CRANE or the treatment couch being inaccurately controlled.
The original method proposed by Dogan et al (2000a, b) and Sethi et
al (2000) to achieve this involved extending the target by a pseudo target
region. In follow-up papers, Leybovich et al (2000a) have improved IMRT dose
distributions by creating pseudo structures which are flagged as regions in which
dose should be kept to a minimum. These have improved IMRT plans of the
posterior neck and thoracic spine. Leybovich et al (2000b) have shown that there
are alternative ways to create the second-phase treatment plan (figure 2.3). The
following were considered:
(i) the addition of an adjacent small target extension instead of a full-width
target extension;
(ii) the creation of a small pseudo target some distance away from the original
target;
(iii) the creation of a pseudo target outside the irradiation volume.
They showed that each of these procedures creates a second-phase treatment
plan whose set of abutment slices lie in different longitudinal planes to those
without the target extensions. They then showed that the two-phase irradiation
technique, even with precise positioning of the phantom with respect to the
irradiation beams, leads to an improved dose homogeneity in the PTV using their
technique. They then went on to show that, when matchline errors are deliberately
introduced, either 0.5 mm or 1 mm, the two-phase treatment technique leads to
an improvement in dose homogeneity to the target compared to the single-phase
irradiation. It was found that the use of a pseudo target remote from the main
radiation volume creates the least change in integral dose. Lehmann and Pawlicki
22
Figure 2.3. Methods of target modification to improve dose homogeneity and to minimize
matchline errors: (a) a simple target extension, (b) addition of a small adjacent target, (c)
addition of a distant pseudo target inside the phantom, (d) addition of a distant pseudo
target outside the phantom. (From Leybovich et al 2000b.)
23
(2003) have also proposed adding tuning structures to give extra flexibility on
IMRT planning.
Dogan et al (2000c) have shown that the use of a pseudo structure positioned
in the posterior neck region can improve the dose reduction to spinal cord when
irradiating concave-shaped head-and-neck tumours. They also showed that IMRT
improved dose distributions compared with 3D, conformal and two dimensional
(2D) treatment plans.
Dogan et al (2002b, 2003a, b) have shown how the use of two pseudo targets
in a CORVUS plan can lead to feathering the match line between two intensitymodulated fields that are required to build up a modulation pattern for a field area
which is bigger than that coverable using a Varian accelerator.
Sethi et al (2001a) have further developed a solution to the problem of
matching tomographic IMRT fields with static photon fields. They make the
observation that in the longitudinal body direction, if static fields are matched
to tomographic fields longitudinally and position errors of the order 3 mm occur,
then hot and cold regions regions of the order 50% can result. This is the familiar
field-matching problem that comes from trying to match two very sharply falling
penumbra. Although the hot spots are significantly large, they do localize in a
very small longitudinal space. The solution proposed was to create a buffer zone
within the target volume, some 3 cm wide, and to arrange that both the IMRT
planning and the static-field planning created a dose gradient in this region of
about 3% per mm with the gradients in reverse directions with respect to each
other (figure 2.4). This can be achieved by specifying a variable dose with
distance prescription for IMRT and by using either a hard or a soft wedge for the
static fields. Then, not surprisingly, when these two fields are just opposed with no
error (exact match) this results in virtually no hot or cold spots. Moreover, if errors
up to 3 or 5 mm occur, the hot and cold spots are still limited to a magnitude of less
than 10% although the spatial area over which they arise can be up to 2 cm (i.e.
across the whole width of the buffer zone). In summary, this paper simply exploits
the fact that ramp distributions match more robustly than do sharp step function
distributions. This is incidentally the fundamental reason why spiral tomotherapy
has an advantage since it inherently creates triangle functions longitudinally.
2.1.3 Energy considerations

Dong et al (2000b) have studied the effect of beam energy in prostate serial
tomotherapy. Plans were created using 6, 10 and 15 MV x-rays on the CORVUS
inverse-planning system and it was shown that the treatment plans at the different
energies exhibited very similar results. This had been shown three years earlier
by Sternick (1997) (see also section 6.9).
24
Figure 2.4. (a) Diagram of the target. The superior target is treated with IMRT and the
inferior target with static fields. (b) A common buffer zone around the matchline is used
in treatment planning for both IMRT and static fields. (c) Beam-edge wedge modifiers
are used in IMRT and static-field plans to obtain the desired dose profile in the buffer
zone. Combined dose profiles in the abutment region were measured for various gaps and
overlaps. (From Sethi et al 2001a.)
2.1.4 Concerns about increased treatment time

The MIMiC IMRT technique generally requires the delivery of substantially more
MU than a conventional therapy. This has implications. Dong et al (2000a) have
studied the effects of induction of secondary cancers from IMRT delivered with
the NOMOS MIMiC system. It was found that the risks were calculated to be
15%, 12% and 17% for 6, 10 and 15 MV for radiation-induced secondary cancers
for a course of prostate treatment of 70 Gy. The more penetrating high-energy
x-rays shorten the machine on-time so reduce the whole-body dose exposure but
the high-energy x-rays produce significant neutron dose equivalent which greatly
increases the estimated risk for radiation-induced secondary cancers. Tao et al
(2002) have shown that IMRT delivers a greater integral dose to the body than
conventional conformal plans. Conversely, Della Biancia et al (2002) have shown
that IMRT can lead to a lower integral dose to the patient if a judicious selection of
beam directions is made. The integral dose for different beam arrangements is a
function of the average pathlength in the patient. Kry et al (2003) have calculated
that the risk of secondary cancers roughly doubles with IMRT.
Mutic et al (2001) have shown that room shielding calculations, reassessed
for IMRT, lead to the conclusion that conventional primary barriers are adequate
for both dynamic MLC and serial-tomotherapy IMRT. However, the excessive
25
head leakage produced by these modalities requires an increase in secondary

barrier shielding. The calculations are based on a careful study of a series of
efficiency factors for the delivery of IMRT. Shielding issues were discussed at
length by Low (2003).
2.1.5 Machine features
The NOMOS MIMiC is known to have features associated with acceleration
and deceleration phases of its vane transit. In particular, when vanes move
from a completely open to completely closed position or vice versa, they bounce
slightly against the end stops and this bouncing can be modelled as a sinusoidal
behaviour with damping. Tsai et al (2000) have modelled this phenomenon with
analytic equations and predicted the variation in dose profiles due to this damped
oscillatory motion. The predictions were compared with experiment favourably.
However, they concluded that the output dependence on the various parameters
only borders on clinical significance and the paper is complex in relation to this.
Zheng et al (2002b) have repeatedly delivered the same PEACOCK plan to
assess whether the leaves of the MIMiC are behaving accurately. It was found
repeated dose point measurements agreed with each other to within 5%.
Hossain et al (2001) have studied the dose output as a detailed function of
the switching frequency for the leaves in the NOMOS MIMiC collimator. It was
found that the output is enhanced as the switch rate increases and therefore every
patient plan needed to be renormalized based on the actual measurements taken
during the delivery of a specified intensity pattern to a phantom.
Huang et al (2001a) have added a tertiary collimator called a Beak Slit
collimator to the MIMiC (Salter 2001) to improve the conformality of dose to
a very small clinical treatment volume. This device truncates the bixel size at
isocentre to 1.0 cm 0.39 cm. Zinkin et al (2004) showed that the use of the
Beak Slit improved tomotherapy compared with the use of the MIMiC alone
with 1-cm-wide slices.
Salter et al (2001) have used repeat CT scans to measure the repositioning
accuracy of the NOMOS TALON removable head frame. Twenty-six repeat CT
scans were used for nine patients. The mean magnitude of isocentre translation
was 1.38 mm during a treatment course over six weeks. Average rotations were
less than 0.5 .
Salter et al (2003) have made a comparison of the dose distribution delivered
to a patient with and without correction for motion as deduced from BAT
measurements. It was found that the mean dose to the tumour could be reduced
from some 97% to 88.3% if BAT corrections were not made. Orton and Tome
(2003) have made a similar observation from planning studies with and without
ultrasound-determined motion correction. For a detailed consideration of the use
of ultrasound in assisting radiation therapy delivery, see section 6.10.5.
Huang et al (2002b) have reviewed the plans for 29 patients with prostate
cancer treated with IMRT and planned using the NOMOS CORVUS system 3.0.
26
Some of these plans included three or four couch movements with gantry rotations
from 240 to 120 clockwise. The plans were compared with conventional plans
and the outcome showed adequate dose distributions to the prostate with sparser
dose to the bladder, rectum and femoral heads.
Woo et al (2003) have described the evolution of quality assurance for IMRT
delivered with the NOMOS MIMiC. This paper is a definitive description of
the various tests both of the system and of individual patient dosimetry that are
advisable when the MIMiC is in use.
Sanford et al (2000) have developed a technique to verify tomotherapy
delivered serially using the NOMOS MIMiC. Dose distributions were evaluated
with a spatial resolution better than 5 mm and errors in absolute dose were
detected within a tolerance of about 6%.
Kapulsky et al (2001) have used the Peacock polystyrene film phantom and
radiation therapy film dosimetry system for comparing planned and delivered
dose distributions for the NOMOS MIMiC MLC system.
In summary, since the development of other IMRT delivery techniques
the MIMiC continues to be widely used in the USA. However, the volume of
technological development has, not surprisingly, slowed down since the major
equipment developments were in the early to mid 1990s.
2.2 University of Wisconsin machine for tomotherapy

2.2.1 Development history
The tomotherapy machine of the University of Wisconsin is a purpose-designed
technology which also uses a MIMiC-like collimator to deliver a single slit of
variable modulation with the variation controlled by the dwell time of the vanes
(figure 2.5). Instead of the double set of 20 vanes, this MIMiC has a single set
of 64 vanes (figure 2.6). The gantry rotates more than once continually as the
patient traverses longitudinally and slowly through the beam. From the patients
perspective, the radiation slit then appears to execute a spiral with fine pitch
(figure 2.7). In spiral tomotherapy the fan-beam thickness (FBT) can be chosen
by the operator to balance the fast treatment time and dose modulation in the
superiorinferior direction. The pitch factor is the couch movement per rotation
in units of the FBT. Typically used values between 0.25 and 0.5 mm result in
advantageous overlap between adjacent helical rotations. The modulation factor
(MF) is the ratio of maximum leaf opening time to the mean leaf opening time in
any projection and can also be user-selected (Kron et al 2004a). IMRT delivery
is complemented by the addition of onboard megavoltage computed tomography
(MVCT). The fundamentals of the device were announced in 1992 and the first
clinical deliveries were in August 2002. A commercial company, TomoTherapy
Inc., has been formed to manufacture and market these machines (figures 2.8, 2.9
and 2.10). To put the developments in perspective, there are at present very few
of these machines compared to the hundreds of MIMiCs in clinical service for
University of Wisconsin machine for tomotherapy
27
Figure 2.5. This photograph shows the helical tomotherapy unit with components labelled.
This unit is the one installed at the M.D. Anderson-Orlando Clinic in Orlando, FL and
represents an improvement in the manufacturability and serviceability as compared to the
University of Wisconsin prototype; the linac jaw system MLC and detector system are
nearly identical to the University of Wisconsin unit. (From Mackie et al 2003.)
sequential tomotherapy and to the large numbers of MLCs capable of performing

IMRT (see section 2.5 for a discussion of the impact of this statement). This
situation may change (Mackie et al 2003b, 2004). At the time of the 14th
ICCR (Seoul, May 2004) four centres had treated patients with the commercial
Tomotherapy machine (Kron et al 2004a). The perspective on the importance
of this device is coloured by the fact that, for 10 years or so, it has been under
development very much in the full gaze of the peer community (unlike the history
of the NOMOS device). The design has also changed over the years; for example,
initially an onboard x-ray kVCT system was planned. Even now when there are
clinical commercial installations, its role is hard to assess because the numbers
are small. This will probably change dramatically in the next few years as its
clinical impact grows.
28
Figure 2.6. Photograph of the helical tomotherapy binary collimator as mounted on the
helical tomotherapy unit. A bank of leaves is visible. Sixteen of the 64 valves are visible on
both left and right. Local high-pressure air reservoirs are visible above and below. (From
Mackie et al 2003b.)
2.2.2 Generation and use of megavoltage computed tomography (MVCT)

images
It is important to distinguish that, in the last four years or so, the developments
at Wisconsin have moved from simulations and laboratory measurements (the
tomotherapy workbench) to the Wisconsin prototype and now to commercial
models. As a result, when reviewing papers, the statements made must be
interpreted in terms of which phase of development they refer to.
Ruchala et al (2000a) have described how MVCT images can be obtained on
the University of Wisconsin tomotherapy workbench (a model of the tomotherapy
system in which the phantom rotated on a spiral assembly in front of a stationary
linac) using the radiation which is also used for treatment. Two major problems
were addressed. The first is that only a limited number of MIMiC vanes are open,
those directed towards the PTV, and so the datasets are necessarily incomplete.
The second problem is that, in order to compute line integrals using the usual
logarithmic attenuation method, it is necessary to know the input fluence to
each line and this is also not a constant because the vanes are clearly moving
to create the modulation. This latter problem was overcome by noting that
there are periods at which the vanes are fully open and these can be used
to normalize the projection data for a limited subset of measurements thus
Gantry continuously
rotates in the same
direction
29
MIMiC with 64 vanes in

single bank
Patient is translated slowly

longitudinally
stop
start
Apparent spiral trajectory in the reference frame

of the stationary patient
Figure 2.7. Schematic diagram showing the principle of operation of helical tomotherapy.
The irradiation of a whole volume is shown. The MIMiC has only one bank of 64 vanes.
The accelerator with this MIMiC rotates continuously around the patient without reversing
direction whilst the patient slowly translates longitudinally. From the perspective of the
stationary frame of reference of the patient, the beam appears to execute a helical trajectory.
permitting reconstruction. Because this is not a full dataset, the contrast resolution
in the reconstructed results is necessarily not optimized. To overcome this, and
also to address the problem of limited coverage of the data, Ruchala et al (2000a)
proposed introducing a number of flashes in which all vanes of the MIMiC are
opened for a very short period of time at a regularly spaced number of angular
intervals and these data used to assist with reconstruction. They showed the results
of this technique on both phantoms and on a German shepherd dog cadaver.
The name given by Ruchala et al (2000a) to the measurements made with the
MIMiC vanes open in a stable position was spairscan. They also comment on
30
Figure 2.8. The cover of the first brochure for the Hi-ART TomoTherapy machine. Note
that, as with the advertisement of CT scanners and MR scanners, the clinical equipment
appears to be little more than a hole in a box, disguising the complexities beneath. Note
the strapline Tomorrow begins with Tomo (courtesy of Professor Rock Mackie).
the technique of reconstructing MVCT images using pure leakage (transmitted)

radiation.
When a MIMiC MLC is used to deliver tomotherapy, the leaves close
completely for a short time within each angular band defining a 1D intensitymodulated beam (IMB). So, for a nearly complete arc treatment, there are 69
such moments (of about 5 ms) in which the MIMiC is doing nothing other than
leak radiation at about 0.3% of the open-field fluence. Ruchala et al (2000b)
have shown that MVCT images can be computed using this entirely leakage
31
Figure 2.9. The University of Wisconsin Tomotherapy Unit with its covers off. (Courtesy
of Professor Rock Mackie.) (See website for colour version.)
radiation (figure 2.11). Thus, no additional dose nor time is taken and the
images are essentially free. It was calculated that, on average, some 20% of
the time the MIMiC leaves are all closed (this clearly depending on the inverse
problem being solved). Thus, when a tomotherapy is set-up to deliver 175 cGy
to the patient, 35 cGy leakage images result. i.e. the images are as if the
treatment would have given 35 cGy with all leaves open but in fact all-leavesclosed delivery had taken place. Density differences above 8% for 25.4 mm
diameter objects and 3 mm air cavities could be observed. Ruchala et al (2000b)
pointed out that these images are not really free because they result from unwanted
leakage radiation which also provides patient dose. The laboratory benchtop test
machine used a NOMOS MIMiC collimator so the 7 cGy image in figure 2.11
is specific to this collimation. Typical MVCT images from the first-generation
TomoTherapy machine were published by Ruchala et al (2002d) and, for the
second-generation machines, images can be created with under 1 cGy due to the
substantially lower leakage (0.3%) of the collimator on this machine (Ruchala,
private communication). The reduced patient dose burden allows use of more
dose with direct imaging techniques instead of the leakage method.
Ruchala et al (2002b) have created MVCT images and, with a 2 cGy scan,
contrast of 3% can be clearly seen and air holes can be resolved at sizes of 1.2 mm
for a 512 512 reconstruction matrix covering the 40 cm field of view. These
images can be automatically registered to kVCT images to further assist with
patient repositioning and verification. Ruchala et al (2003) have demonstrated
32
TomoTherapy System
Figure 2.10. A schematic diagram of the University of Wisconsin Tomotherapy Unit

showing the important portal imaging detector. (Courtesy of Professor Rock Mackie.)
Figure 2.11. Three images of a Rando phantom created by megavoltage computed

tomography using the University of Wisconsin prototype tomotherapy system: (a) image
created with 7 cGy MVCT, (b) image formed with just leakage radiation and all beam
elements closed and (c) an MVCT leakage image using the views collected solely during
the 69 inter-projection regions when all of the leaves were closed. The air cavities in the
phantom are 3 mm in size. These figures illustrate how MVCT images can effectively be
obtained for free. (From Ruchala et al 2000b.)
33
how the MVCT system can overcome image artefacts due to metal prostheses,
patient size being overlarge and to aid planning and replanning.
Mackie (2001) and Olivera et al (2001) have described the developments
in the helical tomotherapy prototype installed at the University of Wisconsin,
Madison. The new Tomotherapy Unit in the University of Wisconsin radiotherapy
clinic has a CT facility that, at low contrast, allows the resolution of objects of
2.5 cm in diameter that are 1% different in density from their background using a
dose of just 3.5 cGy. Such CT images allow the patient positioning to be checked
prior to tomotherapy. Mackie et al (2003a) has provided a detailed overview
review of image guidance for precise CFRT.
Olivera et al (2000b) have addressed the issue of patient movement in helical
tomotherapy. By measuring the movement parameters (shifts), one possibility
to correct the treatment is to reposition the patient using the reverse direction
shifts. However, this is difficult especially correcting for pitch, yaw and roll.
Another possibility is to maintain the patient in the same position on the couch
and modify the treatment delivery. They have created a delivery modification
methodology and presented dose-volume histograms and dose distributions to
analyse the capabilities and limitations of this method. Zhang et al (2004c) has
described the most up-to-date method and this is reviewed in section 6.10.12.3.4.
Zhang et al (2004c) point out that Tomotherapy is specifically disadvantaged in
that it cannot use conventional gating methods nor methods which make the MLC
breath to mitigate motion effects. Fitchard et al (2000) have used canines to
study the effects of mis-registration of the patient in helical tomotherapy.
Kapatoes et al (2003) have used the onboard MVCT images to create a map
of delivered dose using the input sinogram and showed how it compares with the
planned dose.
2.2.3 Clinical application
Olivera et al (2000a) have presented four clinical conformal-to-target and
conformal-avoidance optimization cases for tomotherapy. It was demonstrated
that the complexity of delivery is almost independent of the complexity of the
optimization.
The Wisconsin tomotherapy device is now a clinical reality (Mackie et al
2003b, Kapatoes et al 2001a, b, van Dyk et al 2003, Grigorov et al 2003).
Mackie et al (2002) and Olivera et al (2002) have shown how the University of
Wisconsin tomotherapy system can be used to deliver simple radiotherapy as well
as complex IMRT. It can, in addition to being used for IMRT in a helical fashion,
deliver conventional radiotherapy by topotherapy. Because there is no flattening
filter, the open beam has an intensity that is highest at the centre of the beam and
decreases nearly linearly towards the ends of the beam (Jeraj et al 2004). Hence,
a modulation is required to produce a pseudo uniform beam. Mackie et al (2002)
describe how straightforward this is to achieve. The machine is also being used
to deliver complex whole body irradiation (Mackie et al 2003b).
34
Kron et al (2002) have compared plans created for tomotherapy with IMRT
plans created using the Theraplan Plus treatment-planning system. The advantage
of the former was shown to be improved sparing of normal structures and the
ability to deliver high doses to more than one target simultaneously and/or to
create in-target boost volumes. Grigorov et al (2003) in the same team have
created prostate IMRT plans for the tomotherapy machine.
Welsh et al (2002a, b) have compared tomotherapy, conventional 3D CFRT
and IMRT treatments for a specific case of a tumour close to the spinal cord and
demonstrated that conformal avoidance using tomotherapy could leave the cord
with a sufficiently small dose that future palliative treatment could be possible.
2.2.4 Commissioning issues

Kapatoes et al (2001c) have studied second-order effects in tomotherapy due to
leaf bouncing and leaf latency. They show that although such features are complex
they are dosimetrically insignificant.
Paliwal et al (2002a) have discussed the commissioning of the University of
Wisconsin clinical tomotherapy unit. They pointed out the necessity to accurately
align the MLC with the source and described a method of doing this by making
images in which all the even leaves are open and comparing the images for two
positions of the gantry 180 apart. It turns out that the megavoltage detector
on the tomotherapy equipment has sufficiently good resolution to adequately
reproduce the tongue-and-groove images of the accurately superimposed fields
for alternately even and odd opened MLC leaves. Paliwal et al (2002b) have
presented early experience on the University of Wisconsin tomotherapy unit.
They described acceptance testing and commissioning and compared the system
with another IMRT system. They compared measured and optimized planning
dose distributions.
Balog et al (2003) have presented techniques for commissioning dosimetry
for clinical helical tomotherapy. As helical tomotherapy machines do not
have a flattening filter, this results in a treatment beam with a strong
triangular shape across the transverse direction. This has to be modelled and
measured as a precursor to dose calculation. The helical delivery process also
superimposes inferiorsuperior dose distributions with slight translation offsets so
any modelling error in the longitudinal axis dose profile would become significant
over increasing translation offsets associated with more patient slices treated. As
a result, this profile must also be measured accurately. The lack of the flattening
filter is beneficial for many reasons. In particular, the beam output is more than
twice what it would be along the central axis without the filter and, since all points
at some stage in the tomotherapy delivery are receiving radiation along a central
axis, this decreases the treatment time considerably. Finally, the effect of adjacent
open beam elements has to be calibrated into the dose calculation. Jeraj et al
(2004) have presented the detailed radiation characteristics of the machine.
Tomotherapy using a 60 Co source
35
2.2.5 Verification of MIMiC and University of Wisconsin tomotherapy

Paliwal et al (2000) have created a novel form of verification phantom for spiral
IMRT. A solid-water cylindrical phantom 30 cm in diameter and 15 cm in height
was machined to create an Archimedian spiral cavity in which radiographic or
radiochromic film could be located. The arc length of the spiral cavity was
89 cm. The advantage claimed is that the film is then always at right angles
to the direction of arrival of radiation and also can sample the 3D space with high
resolution, something which is not possible with BANG-gel dosimetry. Paliwal et
al (2001) have reviewed the use of the spiralogram for verifying the accuracy of
tomotherapy. Richardson et al (2003) have provided further details of the spiral
phantom which allows a 3D film-based verification of IMRT without recourse to
the use of multiple orthogonal films. The ADAC PINNACLE TPS was adapted
to predict the spiralogram for comparison. Additionally thermoluminescent
dosimetry (TLD) validated the spiral film method.
The way the system works in practice is that whatever beams are planned
onto a patient are then replanned onto the circular phantom and delivered. Then
measured dose distributions are compared with those which are sampled along the
Archimedian spiral from the planned dose distribution. The space involved is a
so-called spiralogram. The film was calibrated in the usual way. An example was
given in which it was claimed that the measured spiralogram closely followed the
predicted spiralogram.
Lu and Mackie (2002) have developed a method to track the movement of a
patient in sinogram space. They then proposed to use this tracking to reconstruct
CT and single photon emission computed tomography (SPECT) images as if the
patient movement had not taken place. They then extended the notion to explain
that the concept could also be used to track intrafraction organ movement during
tomotherapy.
Sen et al (2000) have devised a technique for recording the modulated
intensity patterns of a MIMiC on a film as the gantry rotates which acts as a
good quality-assurance tool for verifying the planned treatment.
2.3 Tomotherapy using a 60 Co source

Gallant and Schreiner (2000) are developing a tomotherapy apparatus using 60 Co
sources. A large number of 60 Co sources are mounted in a ring gantry similar in
design to a standard CT device and an optimization algorithm based on a modified
genetic algorithm determines which beams to turn on, when and for how long, in
order to deliver the desired 3D dose distribution.
Schreiner (2001) has discussed the practical potential for IMRT with 60 Co. A
first-generation tomotherapy test jig with a rotatetranslate stage was constructed
in which a phantom moved through a 1 cm2 60 Co pencil beam. The beam
intensity was then modulated either by turning the 60 Co unit on and off as
the phantom was stepped through the beam or by varying the velocity of the
36
phantom as it moved through the pencil beam. Early experiments showed

that this could generate a fluence modulation for a 60 Co beam. A series
of fluences with modulation have been applied to a phantom containing a
polyacrylamide gel and a treatment plan for 60 Co IMRT generated with the
MDS Nordion Theraplan Plus treatment-planning system was delivered to the gel
successfully. Kerr et al (2001) showed that measurements and simulation agreed
to 2.5%.
Barthold et al (2002) have described CoRA which is a design for a new
60 Co radiotherapy arrangement with multiple sources. This is a feasibility study.
The apparatus designed comprises five 60 Co sources mounted on a U-arm which
can pivot relative to a patient lying on a couch. The pivotal motion is such that
the sources remain focused to an isocentre and describe arcs within the target
volume. At the same time, the couch may take up different couch twists. The
combination of the U-arm gantry movement and the couch twist can be modelled
as an equivalent set of conventional gantry rotations and couch twists in order to
study the treatment planning.
Barthold et al (2002) modelled the 60 Co source using the Bortfeld pencilbeam technique and used the VIRTUOS module inside the VOXELPLAN
treatment-planning system to model four patient cases. The overall outcome of
the study was to show that the 60 Co-based apparatus could generate conformal
distributions almost as acceptable as those produced by a radiotherapy linac.
The idea is that the 60 Co device could be of use to developing countries. They
therefore laid to rest the potential criticism that the large source size of a 60 Co
system could lead to deteriorated dose distributions.
Warrington and Adams (2001, 2002) have compared, using the HELAX
treatment-planning system, IMRT plans for a linear accelerator and tele60Co
beams. It was found that for thyroid, parotid, maxillary antrum, brain and breast
tumours the 60 Co-IMRT modelling was satisfactory compared with the linear
accelerator modelling. Less favourable sites were the deeply seated lesions in
the chest and pelvis.
Singh et al (2002) and Iyadurai et al (2003) have constructed a manual
prototype MLC with lead for the TH-780C tele60 Co unit. It consists of 15 pairs
of lead leaves with projected leaf width of 8 mm at the isocentre producing a
maximum field size of 12 cm 12 cm. The leaf thickness was 8 cm and a tongueand-groove (TG) design was used to reduce the interleaf transmission to 6% of
the central-axis dose. The 60 Co beam characteristics such as the depth-dose, the
beam profiles and the output factors were found to be very similar to those for
conventional collimation and the MLC will allow beam shaping.
Kim and Palta (2003) have conceptualised/designed a system for using a
plurality of 60 Co sources arranged in a grid matrix for delivering a 2D intensitymodulated beam (IMB) (figure 2.12). The sources may each be separately racked
towards and away from the patient and the radiation intensity in each cell is thus
determined by the inverse-square law change in fluence which will determine the
dynamic range of intensities.
Tomotherapy with an MLC
37
Figure 2.12. A figure from the patent for performing IMRT with 60 Co sources. The
sources (e.g. 108) are housed in bixellated apertures defined by septa (e.g. 104). They can
be moved to and away from a plane (112) so varying the intensity of radiation over that
plane (shown in the lower part of the figure) by simple application of the inverse-square
law. (From Kim et al 2002.)
It is important to note that these 60 Co-based studies and devices are either
concepts or one-off developments and do not (as yet) rival other IMRT delivery
techniques. They are included for completeness and for their potential.
2.4 Tomotherapy with an MLC

Achterberg and Mueller (2001) have developed a new multi-focal static
tomotherapy system which incorporates an MLC and table movement. This
conceptual study has been made using the Monte Carlo BEAM code and the
ADAC PINNACLE3 treatment-planning system and they presented an optimized
design of the static tomotherapy device.
Deloar et al (2004) have produced a design study for kilovoltage
tomotherapy in which a microMLC will be attached to a CT scanner allowing
both imaging and treatment with the same machine. The dose rate is lower than
for 60 Co tomotherapy and, as yet, the machine is only a concept.
38
2.5 Summary
The early years of clinical IMRT (199497) were dominated by the use of the
NOMOS MIMiC. The technique is still widely used particularly in the USA,
though not so much in Europe. MLC-based technology for delivering IMRT is
widely considered to be a front runner rival and may have soon surpassed the
use of the MIMiC at least on a pure number count of completed treatments.
Spiral tomotherapy is in its infancy and like many other technologies that rival
more conventional equipment for delivery, possibly faces an uphill struggle
for adoption. In this sense, whilst it may present significant technological
advantages, it shares this problem with technologies such as proton therapy and
the development of the Cyberknife, technologies believed to display advantages
but clearly in the short-term expensive and numerically inferior to their main
rivals. Against this background, a familiar Catch-22 arises of the difficulty of
obtaining proof of the improvement for lack of the very equipment trying to be
justified. We now turn to the MLC-based IMRT technique, the technique with the
fastest growth in recent times.
Chapter 3
Developments in IMRT using a multileaf
collimator (MLC) (physics)
The history of the MLC goes back at least to the patent by Gscheidlen in 1959
and the first commercial MLCs appeared in the mid to late 1980s. There are
detailed reviews of the history, development, design and performance of MLCs
by Webb (1993, 1997, 2000d). The major manufacturers and distributors of
commercial MLCs, with a 10 mm leaf width at isocentre, are Elekta, Siemens and
Varian. Many papers have detailed the dosimetric performances of each. Galvin
(1999) and Bortfeld et al (1999) have provided useful reviews with tables of MLC
properties. Huq et al (2002) have compared all three commercial MLCs for the
first time using precisely the same criteria and experimental methods (figure 3.1).
The different designs and positioning of the MLCs inevitably lead to performance
differences. However, it was concluded that no single MLC was superior and that
user choice is necessarily not just determined by radiation characteristics.
The Elekta MLC replaces the upper jaws (figure 3.2); the Siemens MLC
replaces the lower jaws; and the Varian MLC is a tertiary add-on at the patient
side (figure 3.3). The position of the MLC varies with respect to target and
isocentre. The Elekta and Varian MLCs have rounded leaf ends but move on a
plane. Conversely, the Siemens MLC has straight leaf ends and moves on the
arc of a circle forcused at the source. The Elekta MLC has a single-stepped leaf
side whereas the Siemens and Varian MLCs have double-stepped leaf sides. Key
observations of Huq et al (2002) were:
(i)
(ii)
(iii)
(iv)
(v)
the Elekta MLC generated the smallest collimator scatter,

the Varian MLC had the sharpest penumbra,
the Siemens collimator had the smallest radiation leakage,
the Elekta MLC had the deepest tongue-and-groove (TG) underdose,
the Varian MLC had the largest stepped-edge effect for a 45 edge.
However, these are statements somewhat out of context and a full study of the
paper is needed to give the complete picture. Some properties of MLCs are given
in table 3.1.
40
Figure 3.1. (a) A schematic diagram of the relative positions of the jaws, the 10-mm-wide
multi-leaves and bottom surfaces of the jaws to isocentre for three different MLC systems
(Elekta, Siemens, Varian). The distances are given in centimetres. (b) A schematic
diagram of the end-on view of various leaves from different manufacturers showing the
differences in leaf design that affect intra- and inter-leaf leakage. Other diagrams in this
paper show the variation of head scatter with field size, the variation in penumbra for the
three collimators and radiation leakage patterns for the three collimators as well as the
variation in tongue-and-groove effect for the three collimators. (From Huq et al 2002.)
Name of
MLC
Number of
leaves
Maximum
field size
(cm2 ) at
isocentre
Elekta (1)
Integrated
MLC
80
40 40
1 cm
Elekta (2)
Beam
Modulator
Millennium
MLC-120
80
16 22
4 mm
120
40 40
Millennium
MLC-80
Millennium
MLC-52
m3
80
40 40
Central 20 cm
of field: 0.5 cm
Outer 20 cm of
field: 1.0 cm
1 cm
52
26 40
1 cm
52
10 10
3.0 mm,
4.5 mm and
5.5 mm
4.0 mm
1.6 mm
5 cm
5 cm
1.4 cm
2.5 cm
4.5 mm
5.5 cm
2.5 cm
Manufacturer
Varian
BrainLAB
Radionics
MRC
Leibinger/Siemens
MRC/Siemens
3D Line (Wellhofer)
MMLC
Mini MLC
62
80
Moduleaf
Mini MLC
80
48
10 12
7.3 6.4 on a
Siemens
12 10
11 10
Direx
Acculeaf
72
11 10
Leaf width
at isocentre
Leaf travel
Leaf
height
32.5 cm
(12.5
overtravel)
11.0 cm
overtravel
19.5 cm
Max leaf
speed
Weight
None
specified
3 cm s1
50 kg
Suits which
accelerator
Focus
Elekta
Single focus
Elekta
Tilted
2.5 cm s1
Varian
Single focus
2.5 cm s1
Varian
Single focus
2.5 cm s1
Varian
Single focus
1.5 cm s1
35 kg
Varian
Single focus
7.0 cm
9.0 cm
2.5 cm s1
1.2 cm s1
38 kg
40 kg
All
All
Single focus
Parallel
7.0 cm
8.0 cm
2.0 cm s1
1 cm s1
39.7 kg
35 kg
All
1.5 cm s1
27 kg
Single focus
Double
focus
Two sets of
leaf pairs at
90
Table 3.1. Properties of commercial MLCs in Summer 2004.
41
42
Figure 3.2. The Elekta MLC. (From Elekta website.)
The MLC can deliver IMRT in several modes. Historically, the first to be
developed was the concept of linear addition of static fields of different shapes
(Boyer 2003). This is referred to by many names including (i) step-and-shoot, (ii)
stop and shoot and (iii) multiple static fields (MSFs). In principle, the subfields
needed are simply a set of patterns summing to the full modulation and formed
by methods which are well known. In practice, complications arise because
of the need to factor in leakage radiation, head scatter and to take account of
the MLC geometrical constraints, such as that, for some manufacturers MLCs,
interdigitation is forbidden (figure 3.4). These are known as hard constraints
because they cannot be violated. Historically, the determination of the leaf
patterns was separated from the inverse planning (because the planning techniques
were developed before methods for IMRT delivery were available). Hence, there
was a period in which substantial fixes were made to account for the physics
of the delivery through collimation that was not perfect and with pieces of metal
43
Figure 3.3. A Varian accelerator and its MLC used for field shaping and IMRT (from
Varian website).
that were forbidden to take up certain positions. Now there is a trend to take
all this into account at the time of planning so that what is planned is genuinely
what is delivered (see section 6.6). However, not all MLC-based planning and
delivery has yet reached this stage. The determinators of leaf patterns are known
as interpreters or sequencers.
Boyer et al (2000) have reviewed IMRT with dynamic multileaf collimation
(figure 3.5). They described leaf-setting algorithms, leaf-transmission effects,
leaf-position calibration, leaf-position offset corrections for the tongue-andgroove effect and dose-rate effects. They concluded that understanding the
44
Figure 3.4. This illustrates interdigitation using a plastic model MLC. The MLC comprises
nine leaf pairs with alternate pairs shown in blue and yellow. The tongue-and-groove
regions show in green. Some leaf pairs are here arranged to show interdigitation. (See
website for colour version.)
technology is essential for its successful application. Keall et al (2003) and

Keall (2004) have provided the most thorough up-to-date review of dynamic MLC
IMRT.
In this chapter, we follow a logical sequence of presenting the most recent
developments in the use of an MLC for IMRT. We begin with the description
of interpreters or sequencers which convert planned modulated beams into leaf
patterns. Then we consider leakage which leads naturally to a discussion of
how the delivery features of the machine affect the interpretation. This leads
into a discussion of dose-calculating algorithms. Then there is a discussion of
features of the MLC-based IMRT techniques before moving on to dynamic MLC
delivery, combined therapies, IMAT and compensators for IMRT. The issue of
the leaf width of MLCs appropriate for IMRT is discussed and grouped together
with a description of new microMLCs. Finally, there are two sections on MLCbased IMRT verification. The first is concerned mainly with verifying individual
treatment. The second is more concerned with quality assurance of the MLC
equipment. However, many aspects of these latter two overlap and sometimes the
placing of material could have been done differently.
45
Time t
Trailing leaves
Leading leaves
Direction of
leaf motion
Intensity
Distance x
Figure 3.5. Schematic diagram to illustrate the principle of the dMLC technique. The 13
rows represent the positions of the leading and trailing leaves at 13 different times in the
delivery of a 1D profile from just one leaf pair. The leaves move from right to left starting
with their positions as shown in the 13th row and ending with their positions as shown in
the first row (for convenience it is assumed here that the leaves can touch; this is not always
possible in practice). The full pair of lines join the midpoints of the leaf ends and constitute
the trajectories of the leading and trailing leaves x(t) or t (x) (these diagrams are shown in
many orientations in practice and this does not matter). It is a simple matter to construct the
intensity profile I (x) from such a diagram since, considering only primary radiation, the
primary intensity I (x) is the difference in time of arrival of the trailing and leading leaves;
i.e. I (x) = ttrailing(x) tleading (x) where intensity and time are considered in the same
units (they are proportional for constant accelerator fluence rate). Two sample intensities
are shown by small vertical bars at two positions x and the whole profile is shown below
the trajectory diagram.
46
Figure 3.6. These are the procedures required to convert fluence to trajectories (F2T).
(From Keall et al 2003.)
3.1 New sequencers/interpreters

3.1.1 General; dynamic IMRT sequencers with hard MLC constraints on
leaves and jaws
There are numerous interpreters which convert IMBs into MLC leaf profiles.
That due to Stein et al (1994), Svensson et al (1994) and Spirou and Chui (1994)
provided an idealized solution for one leaf pair with no machine constraints and
for dynamic delivery. It simply stated that when the gradient of the fluence was
positive, the leading leaf should move at the maximum velocity and when the
gradient was negative, the trailing leaf should move at the maximum velocity.
In both cases the complementary leaf provides the modulation. When there are
multiple leaf pairs, TG underdose may be removed by leaf synchronization (Van
Santvoort and Heijman 1996). The result does not, however, take any account
of the minimum-gap constraints of the Elekta MLC, which has the strictest
constraints of all the commercial MLCs. In this sense, these solutions are too
limited. Convery and Webb (1998) provided a solution which does account
for machine constraints by using both leaves and jaws to modulate the field for
fluence patterns on coarse grids with uniform-across-bixel fluences. Interpreters
are sometimes described as F2T (figure 3.6) (fluence to trajectory) in contrast
to T2F (trajectory to fluence) (figure 3.7) describing the computation of actual
fluence from the trajectories (Keall et al 2003).
Kuterdem and Cho (2001) have provided a new interpreter which works
similarly but for a continuous fluence distribution in which there are gradual
New sequencers/interpreters
47
Figure 3.7. These are the procedures required to convert trajectories to fluence (T2F).
(From Keall et al 2003.)
changes across the field. This also allows generation of zero primary fluence
as well as using both X- and Y-jaws effectively. Kuterdem and Cho (2001)
formulated the problem of optimizing the time delivery of dynamic IMRT
including the machine gap constraints and firstly attempted a MATLAB solution
using the Optimization Toolbox 2.0. With gap constraints temporarily switched
off, the solution replicated the well-known three 1994-papers solutions (see
earlier). However, when gap constraints were switched on, the problem possessed
local minima and was too sensitive to initial conditions. The solution also took
approximately one day of computer time for a problem of leaf sequencing nine
leaf pairs each with 41 control points.
They therefore started again as follows: Firstly the three 1994-papers
solution was adopted. TG effects were removed by synchronization. This creates
gap violations. So an iterative process minimized reduction of leaf speeds as the
trajectories were modified to remove gap violations. This (somewhat complex
procedure) used look ahead steps (as did Convery and Webb [1998]) to avoid
overblocking by diaphragms.
A test pattern of one IMB from a nine-field prostate plan with 2 mm
increments was used. It was shown that the use of the algorithm gave a rms
agreement between delivery simulation and prescription better than 1 MU. When
no secondary blocking was applied (or just use of Y diaphragms), the match was
less good. The full algorithm operated in almost real time.
3.1.2 Sequencing multiple-static MLC fieldsclusters
The alternative (to the dynamic) technique for using an MLC to deliver IMRT is
to deliver the fields by step-and-shoot in which the beam is off between delivery
of field components. The classic method (figure 3.8) to decompose modulations
48
Figure 3.8. This figure shows the classic method of leaf sequencing for the 1D profile
shown left in (a). The panel (a) indicates the means by which the sequence of leaf ends is
determined and then panel b (right) indicates the resulting leaf sequence as a function of
MUs using the sliding-window technique. (From Boyer 2003.)
Figure 3.9. Optimized and clustered profile with four and seven clusters respectively. The
cluster profile on the left has four clusters and no in-field minima, every cluster being a
single segment. The cluster profile on the right has two in-field minima, seven clusters
leading to nine segments. (From Bar et al 2001b.)
was worked out over 10 years ago (Boyer 2003; see also reviews by Webb 1997,
2000d).
Bar et al (2001a, b) have developed a new technique to sequence the fluence
increments required to generate a beam profile in the MSF technique (figure 3.9).
The sequencer combines features of clustering and smoothing to segment the
IMBs. The outcome is that the fluence increments in the delivery are nonuniform. The sequencer is included in the iteration stage that actually generates
the IMBs and is not applied a posteriori. It takes account of the (Elekta) MLC
49
machine constraints. It was designed to be robust. The work was set against
the background of the view from Que (1999) that no particular sequencer (up
to that date) is universally preferable and also set against the views of KellerReichenbecher et al (1999) that combining sequencing and IMB generation as
one process did not improve the treatment plans. They applied this to eight
mathematical profiles and two prostate IMBs. They concluded that this leads
to treatment plans with fewer field segments and a high MU efficiency. The
conformity is a little better than the Bortfeld (equal-fluence-increment) algorithm,
which as the authors state, is an inherently obvious outcome of using non-uniform
increments.
Wu et al (2001f) have described a new algorithm for segmenting static
step-and-shoot IMRT fields. The work describes a method for converting an
idealized beam-intensity distribution to a deliverable sequence of static MLC
segments. The first stage of the process is a multilevel approximation which
groups the intensity values into minimum K -clusters with respect to a predefined
approximation criterion. What this means is that the intensity levels are assigned,
without change, to a set (cluster) up to a fixed total number of clusters; this is
the value K . The second stage of the algorithm then predetermines the level
that should be set uniformly within each of the cluster so as to minimize the
discrepancy between the original intensity and the multilevel approximation. This
technique is well known in the optimization literature and is known as the Kmeans clustering algorithm. It has already been shown by Evans et al (1997) that
this technique minimizes the total squared error in the problem.
The second part of the interpreter developed by Wu et al (2001f) is to start
with the optimal multilevel approximations and to decompose the K multilevel
approximations into L MLC deliverable shapes. This is done using a graphical
user interface. The first segment to be delivered for each gantry angle is the one
that covers the largest field to be irradiated allowing treatment verification. The
authors show that the outcome of K -means clustering is to distribute the errors
in fluence delivery over a wider geometrical area so that, whilst the total error is
reduced, the effect of the error is spread more widely in space.
Much of this paper is taken up with discussing alternative techniques, in
particular that of Evans et al (1997) who showed (with application to breast IMRT
with high spatial resolution) that the use of a uniform incremental value was
optimum. The authors here argue that they could not use this technique because
the large bixel size, used for prostate radiotherapy would lead to a histogram of
well under 100 values, not amenable to study by the technique of Evans et al
(1997). They also give reasons why the technique of Galvin et al (1993) is less
relevant. They comment on the proposal of Webb (1998a, b) to use a forcedbaseline configuration with subsequent leaf sweep. Finally they discuss in great
detail the physical limitations imposed by the Elekta MLC and, in particular, the
way in which jaws and leaves must be combined to create a minimum number
of segments. The features of the Elekta MLC are built into their decomposition
scheme. They comment that MLC rotation can sometimes be advantageous in
50
particular because it leads to a loss of superposition of cold slices due to the TG

effect (see also section 3.4.2).
The number of clusters and therefore the number of segments depends
critically on the error tolerance which is set during the initial clustering routine.
They show, not surprisingly, that as this error tolerance is decreased (i.e. tighter
tolerance) the number of segments increases. The technique is in use at the
William Beaumont Hospital, Royal Oak, Michigan. Nioutsikou et al (2004) also
showed the effect of trading number of segments against error tolerance.
Lehmann and Xing (2000) have used the CORVUS inverse-planning system
to investigate the variability of conformality with use of a different number of
intensity levels. The numbers chosen were 2, 3, 5, 10, 20, 50 and 100. Plans
were sensitive to the number of levels when this was below 10 and less sensitive
when the number exceeded 20. The number of segments of the treatment field
goes up with increasing number of intensity levels and so the final decision on
how many are needed has to balance the quality of the plan and the practicality of
the delivery.
3.1.3 Non-uniform spatial and fluence steps
Beavis et al (2001) have studied the effect of optimizing both the intensity levels
and the spatial increments of the step-and-shoot leaf sequencer for delivery of
IMRT. They started by noting that, in general, the planning and delivery of IMRT
are decoupled. Matrices of 2D modulations resulting from inverse planning
were sequenced a posteriori. The first, and still one of the most commonly
used techniques, is the Bortfeld technique of discretizing in space on a regular
grid and based on discrete intensity using equal intensity increments. Provided,
as for the Varian MLC, the pairs of leaves may move independently without
hard leaf constraints, 2D profile sequencing reduces to a collection of decoupled
sequencing tasks for 1D fluence profiles. Hence, Beavis et al (2001) concentrated
on the solution of the 1D problem.
The rational for their work is to observe that the use of integral multiples of
units step widths and integral fluence increments can compromise the accuracy of
delivery. They postulated that using non-uniform intensity increments and nonuniform spatial increments will fully utilize the technical capability of the MLC
delivery and result in more accurate treatment.
There are two features to their algorithm. The first is to create non-uniform
step widths by finding a set of steps such that the difference between the largest
and the smallest fluence within the step is less than some specified value. This
specified value is increased until the number of steps collapses down to the userdesired number of steps. The second aspect of their algorithm is to select nonuniform intensity levels to minimize a cost function which is a sum-squared
difference between the discretized and the continuous profile at all the points
specified by the previous spatial discretization. They then investigated four
solutions as follows:
51
(i) In the simplest optimization (so-called level zero), the intensity levels and
spatial increments were uniform, namely the solution proposed by Bortfeld
et al (1994).
(ii) Optimization level one allows the intensity levels to be unconstrained but
constrains spatially to a regular grid.
(iii) Optimization level two is the reverse. Intensity levels are constrained to be
incremented equally but the step widths are optimally selected.
(iv) Optimization level three allows both intensity levels and step widths to be
optimally selected.
They observe that only optimization levels zero and two in which the
intensity levels are constrained to be incremented equally can be delivered by a
classic step-and-shoot algorithm. This they did. They implemented optimization
levels one and three using their own algorithm (Beavis et al 2000b). The study
concentrated on just one profile and, not surprisingly, showed that optimization
level three gave the least error between ideal and delivered fluence maps. No
corrections were made for leakage or known dosimetric transmission effects. It
was found that the most significant improvement in agreement was gained by
allowing optimized step widths alone. They commented that this might lead to
narrow segments but they did not think there would be too many, something
clearly to be avoided because of the potential dosimetric uncertainty attached
to introducing geometrically small segments. They also commented that, in
the orthogonal direction, the MLC leaf width itself provides the fundamental
limitation, something not studied in this paper. In conclusion, the new approach
of non-equally spaced fluence levels improved the discretization as well as the use
of non-uniform grid spacing.
3.1.4 Minimizing the number of segments
Dai and Zhu (2001a, b) have presented a new algorithm to minimize the number of
segments in the delivery sequence for IMRT delivered with an MLC. It does this
by selecting the candidate components that result in residual intensity matrices
with the least complexity. The rational for minimizing the number of segments is
that the delivery time for the MSF IMRT mode has three components; beamon time, verification-and-recording-overhead time and leaf-moving time. In
general, the verification-and-recording-overhead time is the largest such time, so
minimizing the number of segments will minimize the dominant component of
the delivery time and thus lead to a delivery time which is shortest. Chen et al
(2004) make a similar study.
There have already been presented a large number of algorithms to sequence
a 2D IMRT matrix into static field components. The best known ones are those
of Galvin et al (1993), Bortfeld et al (1994), Xia and Verhey (1998) and also
Siochi (1999) and these were all compared by Que (1999). Que (1999) found that
the algorithm of Xia and Verhey (1998) most frequently resulted in the fewest
segments but that no single algorithm was the most efficient for all clinical cases.
52
The algorithms all differ because they differ in the rule for selecting the intensity
level of a segment. Therefore, the difference in performance of these algorithms
is caused by these different rules.
The more complex an intensity matrix is, the greater the number of segments
in the delivery sequence will be. Therefore, an algorithm that allows the
complexity of an intensity matrix to decrease most quickly is the one most likely
to produce a sequence with the least number of segments. This is the basis of the
philosophy of Dai and Hu (1999) who proposed such an algorithm for determining
jaw-setting sequences for IMRT with an independent collimator (i.e. no MLC).
The work of Dai and Zhu (2001b) is an extension of that work for the MLC.
The bulk of the paper explains in detail the algorithm by which each segment
is stripped away from a matrix until the residual intensity matrix becomes zero.
The algorithm of Dai and Zhu (2001b) can operate in four different modes
depending on which of the previously mentioned published stripping algorithms
is used for determining the complexity of the residual intensity matrix. The heart
of the MLC stripping algorithm is that they select at the kth segment the candidate
that produces a (k + 1)th residual intensity matrix with the least complexity
(calculated with one of the published algorithms). They do this under two major
situation constraints, i.e. with or without an interleaf collision constraint.
To evaluate the efficiency of the algorithm and the influence of the different
published algorithms used to calculate the complexity of an intensity matrix they
tested 19 clinical IMBs generated using CORVUS version 3.0 revision 10. They
used two indices to measure the efficiency of the algorithms. The first was the
number of segments (the so-called major index) and the second was the total
relative fluence in the sequence generated by the algorithm (the so-called minor
index). When one algorithm generated a sequence with fewer segments than
another algorithm, it was considered more efficient. If two algorithms generated
two sequences with the same number of segments, the algorithm with less total
relative fluence was considered more efficient.
The major conclusions were as follows:
(1) For all beams the most efficient sequences were generated by one, or more
than one, of the four variations of the new algorithm.
(2) The improvement in the efficiency of their algorithm relative to that of the
published algorithms varied from beam to beam.
(3) No single variation of their algorithm was always the most efficient, although
that in which the Bortfeld et al (1994) algorithm for complexity was used was
more often than not the most efficient in terms of the number of segments.
(4) For all beams, every variation of their algorithm was more efficient than the
corresponding published algorithm itself.
These conclusions held both with and without interdigitation constraints.
The reason why these conclusions are reached is that the new algorithm
searches a much larger solution space than the published algorithms. They refer
to the work of Webb (1998a, b) who had previously published the algorithm for
53
determining the possible number of combinations of shapes which can deliver a

particular modulation and had previously demonstrated that it was not practical to
find the global minimum. The overall conclusion was that multiple variations of
the algorithm should be implemented in an IMRT treatment-planning system so
that the most efficient sequence would not be missed.
Related to this, Gunawardena et al (2003) have designed an algorithm for
minimizing the number of segments when segmenting a plan generated by the
CORVUS planning system The concept is based on difference matrices which
guide the selection of flat regions from the modulated beam. Luan et al (2003,
2004) have also described a method to minimize the number of segments for a
CORVUS plan, based on map theory. Advantages were demonstrated over the use
of the Xia and Verhey (1998) algorithm and the Bortfeld et al (1994) algorithm.
3.1.5 IMFAST
Step-and-shoot IMRT requires specification of intensity-increment resolution
and spatial resolution. The inherent accuracy of matching a segmented IMB
to a planning-system-generated modulated fluence clearly improves as both
resolutions are increased. However, this gives rise to an increased number of
field segments with adjunct problems of increased leakage, radiation scatter and
possibly fractional MUs. The Siemens product IMFAST provides a number of
segmentation techniques each with adjustable constraints. Potter et al (2002)
created continuous-level IMBs in the PLUNC planning system, then sequenced
these into five or ten segments to create discrete-level IMBs. These were imported
into IMFAST which uses five different segmentation techniques described in
their paper. IMFAST also creates error maps, being the difference between the
imported maps and the segmented maps. The segments were then reimported into
PLUNC and doses recalculated. Quality factors were defined in terms of how
closely the original planning goals were met by the segmented IMBs.
The authors found that the number of segments varied considerably between
the several segmentation techniques. Furthermore, the quality of the plan
deteriorated due to segmentation. The differences in quality between the several
segmentation techniques became less as the number of fluence levels increased.
No correlation was found between the high treatment efficiency (low number of
segments) and high optimization quality. There was also no correlation between
the average intensity-map error and the optimization quality. There was no
consistent association between the total MUs and the total number of segments.
3.1.6 The best interpreter ever?
The subject of leaf-sequencing algorithms to create IMRT fields using the static
step-and-shoot or MSF technique is ongoing and indeed there seems no end to the
flow of algorithms being published. Langer et al (2000) have investigated three
sequencing algorithms for MLC positions used to deliver IMRT (figure 3.10).
54
One of these is new; the other two are the Bortfeld technique and the areal
step-and-shoot technique (Boyer 2003). The three techniques were applied to
a 2D IMB for the prostate and also to an arbitrary map. It was found that the
Bortfeld algorithm delivers the minimum number of time units (MUs) but does not
minimize the number of segments. Conversely, the areal step-and-shoot technique
generates a smaller number of segments but at the expense of increased number
of time units. The new technique presented by Langer et al (2000) reduced the
time units to those generated by the Bortfeld technique whilst simultaneously
reducing the number of segments to below that generated by the areal step-andshoot technique.
The approach of Langer et al (2001a, c) is called an integer search
programme. It was shown, by investigating random maps to remove selection
bias, that this technique yields better segmentations than the Bortfeld technique
and better segmentation than the areal technique in terms of its ability to
simultaneously reduce both the number of segments and the total MUs. No
details of the new technique were given in these two abstracts but Langer et al
(2001b) have described in detail the special integer algorithm devised to generate
a sequence with the fewest possible segments when the minimum number of MUs
are used and results were then compared to sequences given by the routine of
Bortfeld et al (1994) that minimizes MUs by treating each row independently and
the areal or reducing routines that use fewer segments at the price of more MUs.
It is important to reduce MUs because this affects the contribution from
machine leakage and lengthening of each treatment session and also potentially
leads to inaccuracies due to patient positioning errors. It is also important to
reduce the number of segments because, for some machines, the time needed to
switch the beam on and off between segments and to move the leaves can be
lengthy.
The authors argue that an efficient sequence of leaf movements should take
the fewest possible segments to generate a map with the number of MUs that
is uses. It should also use the minimum number of MUs for the number of
segments that it uses. Clearly, if either condition is violated, the sequence can
be improved by either keeping the number of MUs constant and reducing the
number of segments or by keeping the number of segments constant and reducing
the number of MUs.
To date, most algorithms have concentrated on reducing one or other of these
but not both quantities, i.e. it is well known that the algorithm due to Bortfeld et
al (1994) uses the fewest possible MUs but may use more segments than the
condition requires. Conversely, the areal algorithm is found to generate fewer
segments than the Bortfeld technique but at the expense of increased MUs. These
algorithms have usually been benchmarked against each other using random IMBs
and clinical examples. However, their performance has not been gauged against
an absolute standard. The absolute standard would be the minimum number of
segments required to generate a sequence using the minimum number of MUs
and, before the paper of Langer et al (2001b), this had not been found. Langer et
55
Figure 3.10. The figure shows five different ways of decomposing the simple IMB shown
at the top left: (a) shows an arbitrary intensity map with integer entries. The leaves move
along the rows; columns separate the positions at which the leaves can stop; (b)(d) leaf
sequences provided by three algorithms that generate the map; (b) the Bortfeld algorithm;
(c) the reducing algorithm; (d) the integer (bidirectional mode) algorithm; (e) the integer
(bidirectional) with leaf collision constraint algorithm; (f) integer with tongue-and-groove
constraint. Individual segments are numbered sequentially at the bottom and the number
of MUs for each segment is shown above its figure. The total number of MUs appears in
parentheses below each identified sequence of segments. (From Langer et al 2001b.)
56
al (2001b)s study aims to ask if it is possible to reduce the number of segments

without increasing the number of MUs beyond the minimum required. They do
this by designing an efficient integer minimization problem. The problem is set-up
to take account of constraints, namely that no MLC leaf must have any holes, that
leaf overtravel can be switched on or off, that leaf interdigitization can be switched
on or off and that TG effects can be switched on or off. First of all, the integer
programme is run in order to generate the minimum number of MUs required
to generate the intensity maps when no restrictions are placed on the number
of segments. Then the program is re-run with this number of minimum MUs
to create the solution with the minimum number of segments. The sequencing
problem thus yields the minimum number of MUs with the minimum number of
segments. In other words, Langer et al (2001b) solved a problem previously said
to have been unsolved.
They applied their new algorithm and benchmarked it to the Bortfeld and
areal algorithms for, first of all, a simple algebraic case which can be checked
by eye or by hand and secondly for a total of 19 intensity maps generated from
planning IMRT of the prostate.
The conclusions were as follows. The Bortfeld algorithm used on average
58% more segments than provided by the integer algorithm with bidirectional
motion and 32% more segments than an integer algorithm allowing only
unidirectional sequences. In summary, the Bortfeld algorithm was generating
the smallest number of MUs but generated too many segments and the integer
solution could improve on this dramatically. Alongside this, the areal algorithm
used 48% more MUs and the reducing algorithm used 23% more MUs than did the
bidirectional integer algorithm, while the areal and reducing algorithm used 23%
more segments than did the integer algorithm. In summary, the integer algorithm
could deliver a solution with the smallest number of segments and the smallest
number of MUs, thus improving the efficient delivery of static-field IMRT. The
only drawback of this solution seems to be potentially large solution times but
a further advantage is that it is possible to tailor the integer solution to work
within constrained bounds of number of segments and number of MUs. In fact,
this is a multidimensional optimization problem in which the user can have a
large measure of control over the outcome. Li et al (2004a) have shown how
to minimize the number of MUs and the number of field segments whilst also
avoiding leaf-end abutments. With these observations it is tempting to say the
MSF interpreter problem is solved for the optimum sequencing and that there is
no need for further work in this area.
3.1.7 Sequencers exploiting MLC rotation
Beavis et al (2000a), noting that the resolution of IMRT beams is limited by
the width of the elemental multileaf, have developed an algorithm to determine
whether orienting the MLC in a direction other than parallel to one major axis of
the 2D IMB would be preferentially beneficial. Intensity profiles were sampled at
57
1 intervals rotating the MLC about the centre of the field and for each collimator
angle a set of profiles representing the required intensity surface under each leaf
pair was computed with a width equal to that of a multileaf. In general, it was
found that using non-zero collimator angles reduced the difference between the
idealized 2D intensity surfaces and those produced by IMRT using the dynamic
MLC technique.
The MLC can shape the region either with the edges or the sides of leaves
resulting in a different area at the leaf base depending on single or double focusing
of the MLC. Leal et al (2002) have shown significant differences in the treatment
for isodoses enclosing the target depending on this phenomenon but that better
dose distributions were achieved when the double-focused MLC from Siemens
was simulated with a Monte Carlo technique.
Otto and Clark (2002) have designed an interpreter for delivering IMRT
based on changing both the MLC leaf positions and the angle of collimator
rotation. This is perhaps the first entirely novel delivery technique proposed for
some time. The technique aims to overcome limitations in spatial resolution at
right angles to the direction of leaf movement. It also aims to reduce interleaf
leakage and TG artefacts since a variety of leaf orientation is called into play.
Instead of the primary dose to a point being (largely) solely dependent on the
behaviour of two leaves, it becomes dependent on many.
The algorithmic decomposition is an iterative calculation.
Random
orientations and random leaf patterns are offered sequentially and the dose
contribution is found. If this makes the iterative-running calculation of dose
closer to the prescription, then the contribution is accepted. Also contributions
are accepted if the change falls the wrong way (i.e. worse) but within a small
limit. This limit itself decreases as the iterations proceed. So in this way the
technique has similarities with simulated annealing. To further tune the method,
initially just a subset of the possible components are sampled. Then, as the
calculation proceeds, more component possibilities are introduced being between
the other set. Also, each possible component is checked that it can be physically
delivered. So interdigitation and leaf collision can be avoided if required. The
method involves sequential doubling of the number of segments and a gradual
reduction of tolerance margins.
The technique was operated for both delivery of static segments (radiation
off between rotations and leaf movement) and dynamic delivery (radiation on
between movement of collimators and leaves). It was tested for 1-cm-wide leaves
and for 5-mm-wide leaves (figure 3.11).
The method was applied to a five-IMRT-field thyroid case and also to a test
2D sinusoid. Dose distributions were also measured with film for comparison
with calculations.
It was shown that the fit to the desired fluence prescription increased as the
number of components increased; the error tolerance decreased as the iterations
proceeded. Given the stochastic nature of calculations, some were repeated
100 times with different random number seeds and the results fluctuated with
58

(a)
(b)
(c)
(d)
Figure 3.11. Dose conformity for this C-shaped fluence. Maps are shown for the
5-mm-leaf MLC using (a) rotational and (b) conventional techniques. One-cm-leaf MLC
conformity results are displayed in (c) and (d). (From Otto and Clark 2002.)
a Gaussian distribution of fit error. However, the range of solutions was small.
The fit to prescription increased as the range of possible orientations increased.
With a range less than 90 , behaviour was poor. Conversely, behaviour hardly
improved beyond 270 range.
Comparisons were made with conventional IMRT both in static and
dynamic mode and it was shown (for both leaf widths) that accuracy increased
with the use of rotation. Measurements confirmed that both spatial resolution
increased and interleaf leakage decreased with the introduction of rotation.
Hardemark et al (2003) have also exploited MLC rotation in sequencing
IMBs to reduce the number of segments and to reduce errors due to leaf width.
Wang et al (2004) have developed a new sequencer in which the orientation
of the MLC was an adjustable parameter at each gantry position. Then the
collimator angles were chosen so that the maximum effective travel distance of
any MLC leaf pair at each gantry position is a minimum. This was shown to
reduce the number of segments and also the number of MUs leading to an efficient
delivery without compromising conformality.
3.1.8 Comparison of dynamic MLC (dMLC) and multiple static field

(MSF) techniques
The use of an MLC to perform IMRT by changing the leaf pattern with the
radiation continually switched on is known as the dynamic MLC (or dMLC)
technique (or sliding-window technique) (Keall et al 2003). Some of the
advantages of this technique are said to be (i) shorter treatment times and (ii)
59
reduction of dosimetric errors due to fluence stratification in the MSF method.

Disadvantages are said to be
(i)
(ii)
(iii)
(iv)
(v)
less easy to understand,

requires control of leaf speeds,
interruption is not easy to handle,
it is less easy to verify and
more MUs are needed (Keall et al 2003).
Not surprisingly, it is of interest to know how the dMLC technique compares

with the MSF technique. Of course, if the fluence increments are small and
numerous and the same is true of the spatial steps, then the two techniques are
effectively the same as far as the radiation dose is concerned.
Chui et al (2000b, 2001a) have compared the delivery of IMRT using a
conventional MLC by dynamic and segmental methods. The main advantage
of the dMLC technique is that the continuous leaf motion enables the delivered
intensity profile to closely match the desired one, so preserving both the
spatial and intensity resolutions. Conversely, the MSF-MLC method (step and
shoot) resembles conventional treatment with static fields and so is more easily
verified. However, the MSF-MLC method approximates the fluence modulation
by discrete levels of intensity at discrete spatial increments and thus it is expected
that there will be some degradation in dose distribution. The authors set out
to assess the degree of degradation depending on the spatial resolution and the
number of intensity levels.
In their study, they varied the number of intensity levels from 5 to 20 and the
spatial resolution in the leaf travel direction between 2 and 10 mm. Comparisons
were made in terms of dose statistics from the resulting approximated plans. They
also considered the total beam-on times in terms of MUs and the total delivery
time in terms of minutes.
They reviewed the dMLC technique clarifying that the flatter the minimum
slope is, the greater the maximum leaf speed and, thus, the total beam-on time
is reduced. They also pointed out that an iterative procedure was required
to convert the desired intensity (fluence) profiles into working fluence profiles
such that when the working fluence profile was delivered and the effects of
transmission and extrafocal radiation were taken into account, the overall fluence
would approximate the desired fluence profile (see also section 3.2.2). These
effects also took account of the effect of the rounded leaf end.
When implementing the MSF-MLC technique, equally-spaced fluence
increments were studied. This distinguishes this study from those, for example,
of Beavis et al (2001) who considered non-equal fluence increments. To study
variable resolution, they merged multiple 2 mm-wide bins to create pseudo wider
bins, for example, merging three bins of width 2 mm led to simulation of a bin
width of 6 mm. The leaf-sequencing algorithm that they used minimized the total
beam-on time and the particular MSF-MLC method could be viewed as a special
case of the dMLC method with an infinite leaf speed. Although the derivation
60
of the components was arrived at differently, the final arrangement was said to
be equivalent to that due to Bortfeld et al (1994) with the same beam-on time.
The total beam-on time was always less than that from any other MSF-MLC
algorithms. There was no need to apply leaf-collision constraints for the Varian
accelerator. There is no known method to construct a discrete level working
profile to account for the contribution of extra focal radiation and so the authors
used the working profile of the dMLC technique as a starting discrete profile for
the MSF-MLC technique.
For the three prostate patients studied, the overall results of the dMLC plans
were not very different from those of the MSF-MLC plans with 5, 10 and 20
intensity levels. It was concluded that a 10-level MSF-MLC plan with a spatial
resolution of 2 mm in the leaf travel direction was comparable to a dMLC plan.
In contrast, three nasopharynx patients were studied and their conclusions
were similar for each. The modulations contain sharp gradients for the protection
of the cord and brain stem. It appeared that the number of levels mostly affected
the target coverage for MSF-MLC plans whereas the spatial resolution had a more
significant effect on critical organ sparing with a small spatial resolution yielding
a better plan.
For the three breast patients studied, the results were also very similar and
it appeared that a 10-level MSF-MLC plan for the breast was comparable to the
dMLC plan and that further reduction in the number of intensity levels would
not be acceptable. It should be noted, however, that because of the large nonmodulated plateau in breast fields, a 10-level plan probably corresponds to only
two or three fluence segments.
Regarding timing, the beam-on time required for the dMLC delivery was
about 20% more than for the MSF-MLC delivery but the actual delivery time in
minutes for MSF-MLC was 2 to 2.5 times that for dMLC due to the complex
interaction of times taken for moving the leaves. The authors comment on the
relation of their work to that of Budgell (1998) and that of Keller-Reichenbecher
et al (1999) who also studied the fluence increment resolution effect on the
precision of IMRT.
Ting et al (2000b) have compared step-and-shoot and sliding-windows leaf
sequencing for IMRT. Over 130 patients have been treated using one or other of
the two techniques. The conclusions apply to an accelerator in which there are no
beam-on delays between beam segments. They made a number of conclusions
including the following. Step and shoot uses 3050% fewer MUs but takes
between 50100% longer to execute depending on the complexity of the intensity
map. Step and shoot can deliver sharper dose gradients.
Linthout et al (2002) have compared the performance of dynamic versus
step-and-shoot delivery for fields generated with the CORVUS inverse-planning
system and delivered with an Elekta accelerator. They investigated the delivery
of three fields varying from a simple open square to an inverted pyramid. They
concluded that, in all cases, the dynamic delivery was preferred because it needed
fewer field segments (control points), often (but not always) much fewer MUs,
61
was less affected by the calibration accuracy of the MLC and resulted in dose
distributions that more closely matched the calculations. Unfortunately, Elekta
no longer support dynamic delivery.
Naqvi et al (2001a) have invented a technique to turn the step-and-shoot
CORVUS fluence modulations into an equivalent dynamic modulation using ray
tracing. They did this by generating a quasi-dynamic map by splitting the field
segments into many sub-segments. They studied the errors incurred in the staticto-dynamic conversion.
Turian and Smith (2002) compared the MSF segment and sliding-window
IMRT techniques for delivering IMRT treatments for head-and-neck cancer
planned with the CADPLAN HELIOS system and delivered with a Varian
accelerator. It was found that the sliding-window technique delivered a more
uniform distribution than the MSF technique by approximately 5%.
Swinnen et al (2002) have compared static and dynamic delivery of IMRT.
They showed good agreement between both techniques and calculation. Unlike
Linthout et al (2002), it was found that the total number of MUs for the dMLC
technique usually exceeded those for the MSF delivery but that this did not
influence delivery time significantly because there were no interruptions of the
beam required for the former method. King et al (2002) have also compared the
beam-on time for sliding-window and step-and-shoot IMRT treatment plans.
3.1.9 Varian MLC and HELIOS planning system
Andre and Haas (2000) have implemented both the sliding-window algorithm and
the step-and-shoot algorithm for the dMLC technique on a Varian Clinac 2300 CD
accelerator. Plans were prepared using the HELIOS inverse-planning system.
Dirkx et al (2000) have compared two interpreters for the dMLC technique.
The two interpreters are the LMC (leaf-motion calculator) in the inverse
treatment-planning system known as HELIOS and the DYNTRAC algorithm
developed in-house. The DYNTRAC algorithm includes leaf synchronization to
remove the TG effect and also includes collimator scatter effects. LMC does
not account for these and so is less accurate when compared with experiments.
DYNTRAC requires more beam-on time by about 20% because of the inclusion
of the limitations of the MLC.
Bohsung et al (2000) have commissioned the Varian IMRT system for
delivering the dMLC technique at the Charite Hospital in Berlin. They
link the delivery system to the inverse-planning calculator HELIOS in the
CADPLAN treatment-planning system, Before treating patients, a qualityassurance procedure is performed in which each intensity field is transferred to
a rectangular phantom and the dose distribution in a reference plane and the
absolute dose at a reference point is calculated and measured by film and an
ionization chamber. Agreement between the CADPLAN calculations and the
measurements was good and, in addition, BANG-gel dosimetry was performed.
In November 1999, they treated their first prostate patient using the dMLC system
62
and claim this is the first patient so treated in Europe. The quality assurance
procedures at this centre have been described by Boehmer et al (2004).
Bohsung et al (2000) have described BANG gel measurements to verify
a simulated seven-field nasopharyngeal IMRT treatment plan. A number of
performance tests were devised to check leaf-position accuracy and each IMRT
plan has the interesting feature of including some orthogonal non-modulated
fields with a few MUs. These fields are delivered every treatment fraction so
that electronic portal images may be taken.
Myrianthopoulos et al (2001) have investigated the effect of delivery dose
rate on the calculated fluence maps resulting from the HELIOS inverse treatmentplanning system. They concluded that the error between planning and delivery
increased with increasing dose rate and recommended the use of the lowest
possible dose rate for IMRT.
Islam et al (2001) have used the CADPLAN/HELIOS inverse-planning
system to plan IMRT for the 120-leaf Millennium dMLC system. Qualityassurance measurements were made with an ion chamber and with MOSFET
detectors and films. It was found that pretreatment dosimetry showed agreement
between measured and calculated doses to be within 2 mm in spatial dose
distribution and 2% in absolute dose.
Chauvet et al (2001) have used the HELIOS treatment-planning system
coupled to Varian 2300 CD and 23EX linacs equipped with dMLCs with 80
and 120 leaves, respectively. It was found that the treatment-planning software
required considerable tuning before it could be accepted as having converged to
an optimum solution.
Alaei et al (2002) showed that the static and dynamic IMRT techniques
planned with the HELIOS planning system are comparable.
Oliver et al (2002) have compared conformal treatment plans from a Varian
CADPLAN system and IMRT plans from a HELIOS system for complex tumour
shapes located in the skull and close to the spinal cord and demonstrated the merit
of IMRT.
3.1.10 Developments in Elekta IMRT
Elekta have developed IMRT through their International Consortium now
comprising nine research centres (figure 3.12).
Elektas integrated control system, RTDesktop, checks the linac MLC leaf
positions every 40 ms and so effectively provides record-and-verify for every
segment. The dose per MU is better than 1% after installation of the fast tuning
magnetron. It was argued that this leads to a greater confidence in the use of the
accelerator and thus a smaller workload and less resistance to the introduction of
IMRT.
The FasTraQ MG6370 magnetron is the new tuning system for the Elekta
accelerators (see also section 3.4.4). This has improved start-up performance
and a shorter intersegment dead-time, projected to be 1 s. It has dose-per-MU
63
The Elekta International IMRT Consortium members
Figure 3.12. The constituent research centres that comprise the Elekta International IMRT
Consortium. This consortium was formed by just 8 people in 1994 and has grown now to
include many staff from nine centres.
accuracy within 1% for 2 MU or above. It can be retrospectively fitted to an

accelerator head and the MUs per segment can lie between 1 and 1000 and the
beam limit has been raised to 30 000.
3.1.11 Other interpreters
It should now be clear that there are a large number of options for choosing an
interpreter and many considerations when implementing that chosen. These are
summarized in table 3.2.
In this section are collected together statements about a variety of other
interpreters that have been published, some with few details. Gustafsson (2000)
have generated a general algorithm for sequential multileaf modulation which is
able to take into account the restrictions on clinically available MLCs. This is the
basis of the optimization in the HELAX inverse-planning system (now known as
OnCentra).
Litzenberg et al (2002a, b) have studied the effects of incorporating the leafdelivery constraints of a Varian 2100 EX accelerator equipped with a Millennium
120-leaf MLC into the delivery of IMRT. They find that, depending on the
64
Table 3.2. Options/considerations for interpreters/sequencers for MLCbased IMRT

(when selecting or operating an interpreter these are some options/considerations).
Static or dynamic MLC delivery
Interpreting just primary fluence or including scatter, leaf transmission
Along-the-leaf spatial resolution
Across-the-leaf spatial resolution
Account of hard leaf constraints
Minimisation or not of tongue-and-groove underdose
Generation of minimum number of field segments and/or minimum MUs (MSF
technique)
Inclusion of collimator rotation
Use of clustering algorithm (MSF technique). Specification of number of fluence levels
Uniform or non-uniform spatial and fluence steps (MSF technique)
Dose-rate effects
Use of leaves only or jaws-plus-leaves
Effect of leaf miscalibration
Idealised interpretation of just one 1D profile (one leaf pair) or all 2D IMB (all leaves)
A-posteriori interpretation or inclusion in inverse-planning stage
Performance on benchmark 2D IMBs compared with reference interpreter
Generation of fractional MUs
Ability to recomputed delivered dose through link to commercial TPS
parameters set for the control system, dynamic sliding-window leaf sequences can
be produced which do not require beam interruptions nor doserate modulations.
Xia et al (2002c) have constructed a modified areal step-and-shoot algorithm
that overcomes the limited overtravel distance of the Y jaws on a Siemens linear
accelerator. Applying this to a variety of patterns, they showed that, for all cases, it
was possible to extend the field length from 21 to 27 cm. The algorithm essentially
uses the MLC to create corrections between parts of the field in such a way that
the Y jaws are not required to violate their overtravel constraint.
Crooks et al (2000) have investigated several ways in which intensity maps
(IMB matrices) produced in IMRT planning may be sequenced to optimize
particular outcomes. For example, it was shown that two new algorithms could
improve upon those known as the finite-differences method and Robinsons
method. The first new algorithm produced, known as the absolute norm method,
reduced the sum of the IMB matrix entries by the maximum amount at each
stage. It was shown that this results in the smallest number of components.
The second method, known as the two-norm or sum-of-squares method, reduces
the sum of squares of the matrix entries by the maximum amount at each step
and this results in the smallest beam-on time. The so called finite-differences
method in which there is a reduction of a matrix element by the difference of
the adjacent values and Robinsons method in which there is a reduction of the
65
matrix by the removal of the largest rectangular block of numbers, lead to different
results. It was shown, in particular, that beam-delivery time and the number of
segments required were usually found to be complementary parameters, it not
being possible to optimize both simultaneously. These investigations applied to
step-and-shoot implementation of IMRT using static MLC segments.
Crooks et al (2002) have presented three new algorithms to decompose a
152 matrix of integer fluence values into static field components. These were
two techniques which minimized the norm of the residual matrix after subtraction
of a component compared with the norm before subtraction. The norm could
be either an absolute-value norm or a quadratic norm. The third technique was
a slab-dissection method. The implementations did not include MLC machine
interdigitation constraints nor TG effects though it was argued that they could.
The algorithms were tested on a large number of random matrices with different
numbers of levels and different peak values. They were compared with the
performance of the Xia and Verhey (1998) algorithm since this is known to
perform well when compared with others (Que 1999). It was found that the
minimum-absolute-norm method led to fewer segments statistically than the other
algorithms. The same occurred when CORVUS fluence maps were decomposed.
The slab-dissection method was effective in reducing the treatment time. No
single method is consistently best and this is already well known from earlier
studies. Certainly one cannot predict the best by inspecting the matrix. The
authors stated that none of their three methods can simultaneously minimize both
the number of fluence components and sum of the fluence values (treatment time).
However, this is possible with at least one other algorithm (see section 3.1.6,
Langer et al (2001b)).
Agazaryan et al (2001, 2002) have investigated the parametrization of 2D
fluence profiles obtained from inverse planning using scoring indices. They
characterized the profiles using a gradient index, the fraction of the field that is
planned to receive less than the minimal level of transmission radiation (the socalled baseline index) and the efficiency which is the ratio of an average fluence
level and the cumulative MUs. An in-house algorithm was used for sequencing
the profiles. This could take account of the machines hard constraints. They
showed that there is a clear correlation between the deliverability indices and the
accuracy of a delivered dose distribution.
Price et al (2002, 2003) have developed techniques to reduce the number
of segments in the MSF-MLC IMRT technique and thus the overall subsequent
treatment time. This was done by designing a series of concentric ellipsoids
around the target and a gradient was then defined by assigning dose constraints
to each concentric region. The technique was applied to 36 patients and led to an
increase in the sparing of normal tissues. It also led to a reduction in the average
number of beam directions and the average treatment time delivery.
Papiez et al (2001) have presented a new optimal step-and-shoot sequencer
for unidirectional motion of leaves. They compared this technique with the
Bortfeld algorithm and with the areal step-and-shoot algorithm (Boyer 2003).
66
Kamath et al (2003, 2004a) and Dempsey et al (2003) have made an in-depth

mathematical analysis of segmented multileaf collimation. They showed that leaf
sequencing based on unidirectional motion of the MLC leaves is as MU efficient
as bidirectional leaf movement. They showed that if the optimal plan created
by unidirectional motion violated the leaf separation constraints, then there is no
plan (unidirectional or bidirectional) that does not violate the constraints. Kamath
et al (2004b) showed that including the requirement to eliminate TG underdose
increased the MUs a little but still led to the conclusion that the unidirectional
movement gave the most MU-efficient leaf interpretation. These papers are very
mathematical and rigorous.
Boman et al (2003) have presented a mathematical technique for optimizing
leaf velocities for the dMLC technique.
3.1.12 The effect of removing the flattening filter
Most medical linacs use a flattening filter to create an almost flat uniform beam
without modulation. Hence, all modulation studies are based on this flat beam.
Fu et al (2004) have investigated the effect on IMRT of removing the flattening
filter (flat beams if wanted could of course be made by appropriate modulation).
The use of a Siemens, Varian and Elekta MLC for IMRT was simulated as well as
a compensator with and without the use of the flattening filter. Ten nasopharynx
cases and ten prostate cases were studied and the ratio of (i) beam on time, (ii)
total segments, (iii) leaf travel time and (iv) total treatment time with and without
the flattening filter was specifically computed. The total time is of course the sum
of the beam-on time and the maximum of the record- and verify-time and the leaftravel time. It was found that the beam on time typically reduced by about 43%
but the number of segments and the leaf travel times were very similar (just a few
% change). The absolute time reduction varied between about 3% and about 20%
depending on (i) the type of clinical case, (ii) the accelerator manufacturer and
(iii) most importantly the dose rate and maximum leaf speed and the record and
verify time.
3.2 Radiation leakage and accounting for machine effects in

IMRT delivery
This and subsequent sections describe variations on a theme of delivering IMRT
with an MLC. Table 3.3 summarizes methods and factors to consider when
implementing them.
3.2.1 General leakage issues
Lillicrap et al (2000) have pointed out that shielding of radiation from linear
accelerators was historically never designed for IMRT and that the increased
number of MUs needed for IMRT treatments leads to increased leakage doses.
Radiation leakage and accounting for machine effects
67
Table 3.3.
Variations on the use of an MLC for IMRT
Classic multiple-static field (MSF) technique
Classic dynamic MLC (dMLC) technique
Penumbra sharpening (rind therapy) (section 3.3.2)
Dynamic arc therapy (section 3.5)
Combined step and shoot and dynamic therapy (section 3.6)
Intensity-modulated arc therapy (IMAT) (section 3.7)
Modified intensity-modulated arc therapy (e.g. SIMAT) (section 3.7)
Aperture-modulated arc therapy (section 3.7.2)
Simulated tomotherapy (section 3.8)
Features to consider (applying to all or some of the above)
Dose calculation algorithm (section 3.3)
Leakage and scatter; measurement and use of interpreters (section 3.2)
Room shielding (section 3.2.1)
Effect of rounded leaf ends (section 3.2.3)
Dosimetry of small field components (section 3.3)
Monitor unit verification (section 3.3.3)
Field splitting and feathering (section 3.4.1)
Tongue-and-groove effect (section 3.4.2)
Leaf speed limitations (section 3.4.3)
Accelerator stability; delivery of small MUs per segment (section 3.4.4)
There are of course techniques that can reduce the MUs required (e.g. filtering
the IMBs). Webb (2000b) has described how some aspects of the leakage
in IMRT can automatically be included into the wanted fluence through an
iterative feedback process. Essentially the leaf patterns are fitted to the
transmission-corrected profiles. Williams and Hounsell (2001) argue that the
IMRT inefficiencies are probably no more than those associated with wedge
factors. Also the extra time taken to set-up a patient for IMRT and to treat,
together with the effect of decreased MU efficiency, may lead to the total number
of MUs delivered in the working day being much the same with, as without,
IMRT. Hence, daily leakage radiation would be much the same for calculation
of room shielding. (This statement was written at a time when prototype clinical
IMRT generally took longer than conventional IMRT; this may now not be the
case and this rider may no longer apply.)
Ipe et al (2000) have shown that the use of an MLC for the dMLC technique
does not increase neutron production. Measurements of neutrons were performed
at Stanford Hospital with a Varian Clinac accelerator delivering 15 MV photons.
68
Figure 3.13. In this figure, a simple 1D IMB is shown comprising three fluence levels
with equal fluence increments (upper panel). Below this are three possible practical
realizations of the fluence profile. The MLC leaves are shown black and the three
component irradiations are each shaded differently. It can be seen that for three levels
there are 3!=6 possible configurations, three of which are shown. The fluence delivered in
each case would be (slightly) different because of leakage and scatter variations (although
the primary fluence would be the same in each case).
Hover et al (2003) have compared several MLCs and microMLCs and

characterized their leakage radiation noticing differences between manufacturers
and also differences between leaves of any particular MLC.
3.2.2 Factoring in delivery physics
3.2.2.1 Measurement and prediction of leakage and scatter
When an MLC is used for IMRT, it creates patterns with the leaves taking
up different positions (figure 3.13). In order to compute the delivered dose
distributions, a model is needed for x-ray transmission through and scatter effects
in an MLC. These calculations are sometimes described as T2F (trajectory to
69
fluence particularly when in conjunction with dynamic delivery) (Keall et al

2003). A numerical model has been presented by Chen et al (2000a). This
analytic and numerical model was compared against a Monte Carlo calculation
with good accuracy. In order to account for the finite size of the radiation
source, the point-source transmission measurements were convolved with the
distribution of source strength to generate transmission measurements that again
were in agreement with Monte Carlo dose calculations. Arnfield et al (2000) have
presented a method for determining multileaf collimated transmission and scatter
for dynamic IMRT. The purpose of this work was to create the measurements
which allow to take account of transmission when MLCs are used for creating
IMRT using the dMLC technique. It is necessary to know the leaf transmission
as a function of the field size. This is important because the field size varies
during the delivery technique and the transmission can be a significant faction of
the delivered dose. Measurements made by Arnfield et al (2000) were pertinent
to two Varian MLCs, the Mark II, 80-leaf MLC and the so-called Millennium
120-leaf MLC.
The leakage radiation comprises two components: direct radiation and leaf
scatter. The authors measured the direct radiation transmission by extrapolating
the total transmitted radiation back to zero field size. So, for the 80-leaf MLC,
they created a series of fields of length 5 cm with widths 0.510 cm for 6 and
18 MV. They then plotted the leakage transmission across the leaf direction
and generated the usual sinusoidal patterns. They then line-integrated this
transmission as a function of field size and, by extrapolating the graph back to zero
field size, determined that the direct component of transmission was 1.48 0.1%.
The transmission due to scatter from the leaves could then be worked out by
taking this value away from the total transmission. For example, the leaf scatter
of a 10 cm 10 cm field was 0.20 0.01%.
Measurements for the 120-leaf MLC showed that the direct transmission was
1.34 0.03% and the estimated leaf-scatter component was very similar to that
for the 80-leaf MLC, not surprising given that the architecture of the central part
of the MLC is very similar.
Naqvi et al (2000) have developed a dual-source model for the Elekta SL20
linear accelerator equipped with an MLC. This was used to predict output factors
for complicated shaped fields so that MU calculations can be made. Prescription
files from the CORVUS treatment-planning system were read into the programme
and it was found that values for the target dose predicted by the model agreed with
measurement to within 0.3% for any static field segment of arbitrary shaped size
and off-axis position. For the entire delivery, the accuracy was within 0.6% of
measurement using this calculation technique. However, CORVUS, not including
the effects of scatter, overestimated the dose by between 1 and 4% when compared
with measurement. Naqvi et al (2001b) have studied in detail the effects of head
scatter in IMRT. They computed the form of the scatter source representation for
an Elekta SL20 accelerator. They then used ray-tracing methods to compute the
effect of scatter in a measurement plane. Calculations and measurements agreed
70
to within 1%. They were particularly interested in the interaction of scatter with
the complex collimation.
Baker et al (2002) have developed a head-scatter model for the Elekta Precise
treatment unit which is capable of reproducing the measured head scatter to within
0.5% for fields ranging in size from 1 cm2 to 40 cm 40 cm. The model has been
implemented for a PLATO treatment-planning system.
BackSamuelsson and Johansson (2002) have shown that head scatter in
IMRT treatments with the dMLC technique is not only dependent on the jaw field
size but also on the size of the slit opening and the speed of the leaves.
Klein and Low (2000) have noted that, with IMRT treatments taking
approximately five times as many MUs as conventional CFRT, leakage through
leaf junctions becomes an issue. A comprehensive leakage study was performed
for an MLC with both 1-cm- and 0.5-cm-wide leaf systems. They included in the
calculation the effects of patient motion and concluded that interleaf leakage is a
concern for IMRT therapy.
Klein and Low (2001) have observed that interleaf leakage as high as 4%
is effectively magnified to 20% when account is taken of the five-fold increase
in MUs required for many dMLC IMRT treatments. The concern is heightened
for the 120-leaf system with more interleaf junctions. It is expected that patient
motion will moderate the actual effects in practice. Studies were made for a 5mm-leaf-width system (120 leaves on a Varian 23 EX) and for a 10-mm-leafwidth system (80 leaves on a Varian 2300 CD). Films were exposed perpendicular
to the beams. Also a microionization chamber with an effective measuring
width of only 1.5 mm was used to find the location of the maximum leakage
intensity. Wedge-shaped IMBs were created and the intratreatment motion
was simulated by adding translations and rotations and multiply irradiating the
detector. Without movement, the peak leakage was some 4% of the maximum
dose and decreased by about 1% with movement. Additionally, prostate IMBs
computed by CORVUS were delivered under the same conditions including
simulated patient movement. Specifically the parts of the fields where no primary
dose was intended were analysed. Peak leakage dose was around 78% in
these regions without movement and decreased by 12% with movement (details
depending on amplitude of movement and the specific MLC). The effect of
movement was to effectively spread out the leakage in space rather than reduce
its overall contribution.
3.2.2.2 Using leakage and scatter knowledge in the MSF-MLC technique
The problem of incorporating the effects of leaf transmission and head-scatter
corrections into MLC-delivered IMRT is different when the delivery is by a
dynamic technique or by a step-and-shoot technique. For the dynamic technique,
the effects can be incorporated by iteratively adjusting the MLC leaf positions.
However, when IMRT is delivered with the step-and-shoot technique, the a
posteriori segmentation has already determined the fixed leaf positions and these
71
are no longer varied. This segmentation will have been done assuming that
transmission is zero and that head scatter is zero, neither of which are correct
assumptions. As a result, if the dose is delivered with these field patterns, without
any further modification to the fluence values for each segment, the delivered dose
will not be the planned dose because the delivered fluence will not have matched
the planned fluence.
Ting et al (2000a) have developed a new stop-and-shoot leaf-sequencing
algorithm for IMRT which accounts for leaf transmission, interleaf leakage,
scattered radiation and leading-leaf penumbra. Outputs from the leaf-sequencing
method for dMLC control have been compared with measurements made in water
and solid phantoms. The sequencer has been used for more than 100 patients since
July 1998.
Yang and Xing (2002) solved the problem of how to account for leaf
transmission and head scatter in step-and-shoot (MSF) IMRT by the expediency
of adjusting the MUs per segment until the delivered fluence matches the
calculated fluence in a least-squares sense. The algorithm works by adjusting
the fractional MU factors in the leaf-sequence file without considering the two
effects until the delivery with fractional MUs matches the outcome of including
the full physical effects into delivery of the leaf sequences with integer MUs .
This is needed because, unlike in the dMLC technique, the fluence distribution
cannot be continuously varied. The technique is iterative in the same way as that
of Convery and Webb (1997) for the dMLC method.
Yang and Xing (2003) have presented further details. To modify the
number of MUs applied to each segment such that, after modification, the
fluence delivered including the effects of transmission and head scatter is
most closely matched to the planned fluence, they used a downhill Simplex
technique (figure 3.14). In order to implement this technique, they required to
develop a three-source model for head scatter which they verified by predicting
and measuring head-scatter factors for specific small fields and specific IMRT
components, comparing the measured results with the predictions from Monte
Carlo calculations. Leaf transmission was also estimated for a particular Varian
accelerator as about 1.75%.
For both an intuitive test field (a simple 10 cm 10 cm uniform open field)
constructed from five consecutive 2 cm 10 cm segments, and also for an IMB
from CORVUS (part of a clinical prostate plan), they showed that the fluence
maps obtained from their technique, when both transmission and leaf scatter were
included, were in very close agreement with the calculated intensity map without
these corrections. They also showed that measurements made with film agreed
with Monte Carlo calculations very closely. In summary, these two complex
physical effects are entirely taken care of by adjusting the number of MUs per
segment.
Papatheodorou et al (2000a) developed a model to compute the forward dose
for a set of interpreted dMLC field-shapes. This model was an extension of a
primary-scatter separation used for conventional fields. It also took account of
72
Figure 3.14. Illustrating the technique for correcting step-and-shoot fluences (MUs) to
account for head scatter and leaf transmission. The profile is one line through a set
of three measured dose distributions when the fluence components MUs were either (i)
uncorrected, (ii) corrected for transmission or (iii) corrected for both transmission and
head scatter. Monte Carlo calculations are shown for comparison. These agree best with
(iii). (Reprinted from Yang and Xing (2003) with copyright permission from Elsevier.)
73
the rounded leaf ends, the penetration through which becomes increasingly more
significant as the size of the field segments becomes very small. The model used a
single exponential for the head-scattered radiation with a pair of fitting parameters
which could be different to account for the different leaf-end and side geometries.
Calculations and measurements agreed to better than 3% in regions of uniform
dose and to 3 mm in high-dose-gradient regions.
3.2.3 The effect of rounded leaf ends: light-field to radiation-field
discrepancy in IMRT
There is a further phenomenon to be accounted for in the dMLC technique with
the Varian accelerator. Because the leaf ends do not have a perfectly doublefocused form but have rounded leaf tips, there will be an additional fluence
through the tips depending on the position of the leaves in the field and this has
often been described as equivalent to an effective increase in the window width
between opposing leaves over and above the geometrical distance between leaf
tips (figure 3.15). This equivalent shift has sometimes been called a leaf-gap offset
and must be determined for each MLC. The reason is that the effective widening
can be a significant fraction of the variable distance between the leaves. The 50%
penumbra is pushed back in the direction of the tungsten from its position in the
light field and thus it is necessary to reduce MLC leaf gaps in order to obtain
agreement between calculated and measured profiles (Boyer 2003).
Arnfield et al (2000) measured this shift by delivering the radiation through
a sweeping window of variable width. The leaves were set to sweep from one side
to the other of a 10 cm by 10 cm field for a series of fixed separations between
the leaf banks ranging from 0.510 cm. By making the appropriate calculations,
they then determined that the equivalent shift was 0.114 0.004 cm for the 80leaf MLC and 0.088 0.003 cm for the 120-leaf MLC. As a check on their
measurements, it was found that when this was included in the calculation of dose,
the predicted and measured values agreed to within 2%. Arnfield et al (2000) then
backed up their experimental measurements with Monte Carlo calculations.
Kung and Chen (2000a) also pointed out that when using a (Varian) MLC
to perform IMRT using the dMLC technique the radiation-field offset must be
programmed into the machine instructions that drive the MLC. The radiationfield offset is the difference between the position of the 50% radiation boundary
and the light-field boundary. Usually this is small, just a fraction of a mm.
They investigated the dosimetric consequences of using the wrong value and
experiments showed that the dose error sensitivity factor (percentage change in
dose for each mm of radiation-field offset) was in the range 08% mm1 . This
demands that radiation-field offset be no more than 0.5 mm if the dose error is to
be kept to better than 4%.
Cadman et al (2002) have made a study of this phenomenon and come
to the conclusion that an MLC leaf-gap reduction of 1.4 mm is required to
obtain agreement between calculated and measured profiles. This was found
74

S
MLC leaf 1
MLC leaf 2
Figure 3.15. Illustrating the radiation field offset (RFO) phenomenon. The full lines show
the edge of the lightfield defined by two MLC leaves. The dotted lines show connectivity to
the 50% penumbra each side of the aperture. The distance between the two 50% penumbra
points will not equal the lightfield width if the leaf ends are round.
by comparing profiles delivered with single open fields, profiles delivered using
seven IMBs, each delivered separately, that correspond to an IMRT plan and
they also showed that the full dose distribution in an IMRT plan agreed better
with calculation from the ADAC PINNACLE system when this leaf offset was
introduced. They comment that future releases of the ADAC PINNACLE system
will have this discrepancy overcome.
Vial et al (2003) also measured the radiation-field offset and found that better
agreement between calculation and experiment could be obtained by using an
experimentally-determined value instead of the manufacturer-provided value.
3.3 Dose calculation for IMRT

Several studies have addressed issues arising from the types (size and irregularity)
of field shapes generated in MLC-based IMRT.
Martineau and Manens (2001) have shown that there is good agreement
when measuring the depthdose distributions and profiles for very small fields
such as those that will be delivered in IMRT when the following detectors were
compared: diode, a pinpoint chamber, film, a Markus chamber and a diamond
detector. They recommended the use of the diamond detector because of its
good spatial resolution and near tissue equivalence for measuring depth-doses and
Dose calculation for IMRT
75
profiles. For measuring output factors they recommended the pinpoint chamber.
Sanchez Doblado et al (2003) have shown that small-field dosimetry is accurate
based on reference 10 cm 10 cm field dosimetry.
It is well known that in the step-and-shoot IMRT techniques some of the
static field components are very small in area. Cheng et al (2003) have shown
that, when lateral electronic equilibrium does not exist, the output factors for
small fields significantly differ from those that would be predicted using Days
equivalent-square formulation. By making careful measurements, Cheng et
al (2003) produced graphs of output factors as a function of field size under
conditions of electronic disequilibrium.
Knoos et al (2001) have shown that the use of the collapsed-cone algorithm
in the HELAX treatment-planning system led to only marginal changes in dose
distribution from the use of a pencil-beam algorithm.
Cozzi and Fogliata (2001) have compared calculations from the HELAX
treatment-planning system and measurements performed with ion chambers,
portal imaging devices and films. Calculations and measurements agreed to 1%
for output factors and 2% for depth doses and profiles outside high-dose-gradient
regions.
Xia et al (2000b) have studied the influence of penumbral regions appearing
in the middle of the target and dose inhomogeneity within the target introduced
due to patient motion or inaccurate MLC positioning in the step-and-shoot IMRT
delivery. Preliminary results showed dose errors between 2% and 5%.
Phillips et al (2001b) have shown that dynamic IMRT dose distributions
should be modelled as a convolution of a scatter kernel with a continuous fluence
assuming constant velocity leaf motion between control points. If, conversely,
the beam modulation is modelled as a set of static fields, the results significantly
differ from measurements.
3.3.1 Application of colour theory
It is well known in IMRT that the radiation fluence delivered through any one
bixel depends on the state of opening of adjacent bixels. This was first shown
for the NOMOS MIMiC by Webb and Oldham (1996) who provided an analytic
technique for creating dose distributions in the so-called component-delivery
(CD) mode. Essentially the same problem arises in the dMLC technique when
bixels may be identified as being delivered by elemental movements of the
multileaves. Concentrating on any one bixel, the state of the leaves covering
or uncovering adjacent bixels affects the dose distribution through the bixel being
inspected. For example, if one were to inspect a 3 3 panel of bixels and the
central one is open, then there are 28 possible ways to select open or closed states
of the eight neighbours of this central bixel. These 256 possible arrangements
are (in group theory) known as colourings. The 256 arrangements lead to 256
different values for the dose from the inspected bixel.
76
Langer (2000) has analysed the colourings and shown that the number of
independent patterns is less than 256. In fact, it was shown that this number is
51 distinct patterns, if no distinction is made between the method of collimation
using the leaf faces and the leaf sides. If the symmetry between orthogonal
coordinates is broken by collimator leaves whose ends and sides have different
effects on bordering bixels, then the number of patterns increases to 84. However,
it can be seen that both these numbers, being less than 256, lead to a simplification
of the dose calculation. The (fewer number) elementary pattern dose calculations
can be stored and then those which reduce to these through rotations and
reflections, can re-use the same stored dose distributions. In this way, the
calculation of dose distributions in IMRT can be speeded up by a factor of three or
more. Langer (2000) gives examples of this technique. This is possibly now less
necessary given that fast dose calculations are now fairly straightforward without
this simplification.
3.3.2 Penumbra sharpening for IMRT
It has been known for some time that the use of margins, rinds of small spatial
extent but higher fluence than the surrounding fluences, can sharpen the penumbra
in radiation therapy. Sharpe et al (2000a) have performed a thorough investigation
of the dependence of the effect of energy, the width of the rind and the added
fluence. Both calculations and measurements were made at 6 and 18 MV and
measurements were made using both rinds created using different apertures in cut
lead sheets and using multiple multileaf collimated fields.
In summary, their work showed that the 5095% penumbra could be
significantly reduced by the use of high-fluence rinds. The precise fluence
increase required depended on the radiation energy and on the width of the rind
and on the nature of the surrounding tissue, whether water or lung. The effect was
ubiquitous and only precise values depended on energy and material. Sharpe et
al (2000a) went on to show that, for specific examples of a lung tumour treated
with 6 MV x-rays, the margins could be considerably reduced and that penumbra
compensation led to both an improvement in the dose-volume histogram of the
PTV and also simultaneous increased sparing of the spinal cord.
3.3.3 MU verification
Dosimetrists are in the good habit of making routine MU checks of conventional
therapy (non-IMRT). For the delivery of modulated fields (IMRT), this ability
using conventional methods disappears. Other methods are needed and the
following have been developed.
Kung et al (2000b) and Kung and Chen (2000b) have developed an MU
verification calculation for IMRT as a quality-assurance tool for dosimetry.
Huntzinger and Hunt (2000) have developed a manual technique for MU
calculations in IMRT. Lujan et al (2001a) have compared the independent MU
Features of MLC delivery of IMRT
77
calculation from a modified Clarksons method to the CORVUS dose calculation

for a number of different IMRT fields and found that these agreed to within 2.5%.
3.4 Features of MLC delivery of IMRT

3.4.1 Large-field IMRTsplitting the delivery
The Varian MLCs are limited to creating fields of length 14.5 cm in the direction
of the leaf movement when used for the dMLC IMRT technique. This is because
no trailing leaf end may ever extend beyond the jaw position. Yet sometimes
it is useful to be able to create larger IMRT fields, e.g. when delivering a
simultaneous boost. Wu et al (2000c) have solved this problem by a simple
technique (figure 3.16). A large IMRT field is split into two contributions which
overlap spatially. In the overlap region (generally a few cm in length), each IMB
is feathered from half the value at the midpoint of the feathering length to zero
such that throughout the feathering region the sum of the two split IMBs is equal
to the unsplit beam. Wu et al (2000c) showed that there is no need to adjust output
factors. They showed experimental agreement between computation and delivery
using this method. They also suggested that one could split fields whose length
is less than 14.5 cm as this would permit the use of back-up jaws (enveloping the
split fields) to reduce radiation leakage.
Chui (2000) and Chui et al (2001b) have described clinical implementation
of IMRT and specifically techniques for splitting a 2D intensity-modulated
field into two, when the delivered field width exceeds the limit imposed by
the MLC design. Chui et al (2001b) have shown, using an example from a
nasopharynx field, that splitting should be done in the low-intensity region where
cord protection is needed.
Hwang et al (2001) have also developed an algorithm to increase the
maximum usable field size for the Varian dMLC system for IMRT. The IMRT
field length is enlarged to the full length of 27 cm as appropriate to conventional
multileaf collimated fields. The strategy employed was tested on 1000 random
patterns.
3.4.2 Tongue-and-groove effect
The tongue-and-groove (TG) effect is a much studied unwanted underdose caused
by the combined effects of irradiating through extended tongues and grooves
along the sides of MLC leaves. In the delivery of IMRT by the dMLC technique, it
can be entirely eradicated by synchronization. However, unwanted consequences
of synchronization are: (i) increased treatment time and number of segments,
hence increased leakage; and (ii) possible across-the-leaves dose error due to the
requirement for very precise leaf positioning.
Deng et al (2000) have compared, using the EGS4/MCDOSE Monte Carlo
code, the dose distributions using the fluence maps with and without TG effects.
78
Figure 3.16. This diagram shows how to deliver a large-area intensity-modulated field
either directly or by splitting it into two components in which the field boundary is
feathered for each of the two components : (a) profile for the original large field; (b)
simulated film for the large-field intensity distribution. The white cross denotes the
isocentre. Darker areas represent higher intensity; (c) shows the profile for the left
component field; (d) simulated film for the left component; (e) the profile for the right
component; (f) simulated film for the right split field. (From Wu et al 2000c.)
They concluded that, when multiple fields were used and setup errors were
included, there were no distinguishing discrepancies found, indicating that the TG
effect can be neglected clinically (figure 3.17). Deng et al (2001a) conducted a
computational study to show that the effect could, in general, be ignored for IMRT
79
Figure 3.17. A schematic diagram of the tongue-and-groove effect in an MLC: (a) the
design of the MLC tongue-and-groove is to reduce interleaf leakage; (b)(d) schematic
diagrams of two fields and their superposition defined by two adjacent leaves. The region
centred between the two leaves in (d) is underdosed. This effect was shown to have no
clinical significance when multiple individual fields are added together in an IMRT plan.
(From Deng et al 2001a.)
treatments involving multiple ( 5) fields and some modest patient movement.

They came to this conclusion as follows. Firstly, IMRT maps were extracted
from the CORVUS inverse-planning system for two cases. The first case was
an eight-field coplanar treatment of the prostate using an 80-leaf Varian MLC
on a Clinac 2300C/D accelerator delivering 15 MV radiation. The second case
was a nine-field coplanar treatment of the vertebra using an 52-leaf Varian MLC
on a Clinac 2100C accelerator delivering 4 MV radiation. The CORVUS maps
were computed with the TG effect present. Then the MCDOSE Monte Carlo
program was used to computed fluence and ultimately dose, taking into account
leaf transmission and head scatter but ignoring the TG effect (so-called Plan 1).
Alternatively, the files were taken through the MLC2MAP code to include the TG
80
effect as well as the other effects mentioned (so-called Plan 2). Two more plans
were then computed with these two plans blurred with typical patient movement.
Results were shown as dose maps and DVHs for PTV and OARs.
They arranged for a statistical accuracy of some 1% with doses calculated
on a 1 mm grid at isocentre. It was shown that the TG effect produced dose
differences no more than 1.6% when the number of fields were eight or nine and
patient movement was included. The doses were slightly larger in Plan 2 in lowdose regions and covered a smaller area in Plan 2 in high-dose regions. This
small residual effect was considered clinically insignificant. It arises because the
underdose regions do not coincide for each field contribution. This is especially
true with the inclusion of patient movement. In the study, conversely, some singlefield irradiation calculations were carried out showing much larger TG underdose
greater than 10%. They concluded that if the number of fields is small (say 5),
it may be necessary to conduct a study like this to check on the TG underdose.
However, for IMRT with larger numbers of fields, they concluded that we can
finally ignore the TG effect.
Whilst the technique of Van Santvoort and Heijmen (1996) and Webb et
al (1997) can completely remove the TG undersdose for the DMLC technique,
until recently no such equivalent solution had been proposed for the MSF
technique. Indeed Webb (1998a, b) claimed that no such solution could ever
be found via a direct search of the configuration options for the fixed number of
components from a decomposition by the method of Bortfeld et al (1994). Que
et al (2004) have presented a new method to decompose an IMB based on first
applying the technique of Van Santvoort and Heijmen (1996) to create equivalent
(tentative) static-field components and then adjusting them (synchronizing
right leaf positions) to remove the TG underdose in a step-and-shoot delivery.
Applying this to a variety of clinical IMBs as well as random IMBs showed
the problem was solved. However, the price paid is a very large increase in
the number of field components as well as the solution having the smallest leaf
openings (parametrized through their averaged leaf-pair opening [ALPO]) (see
section 3.14.1). Of the more commonly applied decomposition methods, it was
shown that the reducing level method of Xia and Verhey (1998) had the smallest
TG underdose with a realistic number of components.
3.4.3 Leaf-speed limitations

Sohn et al (2000a) have noted that the dMLC technique is most limited when
requested to create shallow-gradient intensity profiles. This is because shallowgradient profiles correspond to the leaves moving at high or maximum speed.
The Varian Clinac 2300 CD accelerator automatically reduces the dose rate when
the leaf velocity reaches its maximum value. The fidelity of generating different
gradients as a function of dose rate and leaf speed were studied. The dose profiles
were linear when the MLC was not moving at maximum speed but artefacts
81
were generated when the leaf velocity limited the dMLC delivery and the profile
became irregular.
3.4.4 Stability of accelerator and delivery of a small number of MUs and
small fieldsizes
It is well known that a feature of IMRT, delivered with the step-and-shoot
technique, is the creation of many small-area field segments and the requirement
to deliver the segments sometimes with very few MUs per segment. This situation
was extensively investigated by Hansen et al (1998). However, the conclusion of
that paper was that the stability of an accelerator required to be measured for each
individual accelerator and that measurements from one could not be translated
directly to experience for another. Hence, Sharpe et al (2000b) have essentially
repeated the work done by Hansen et al (1998) for two accelerators at the William
Beaumont Hospital, Royal Oak, Michigan (figure 3.18). One of these accelerators
was equipped with a slitless flight tube and a second machine, used for IMRT,
initially had a slitted flight tube changed later to a slitless flight tube. Sharpe et
al (2000b) showed that, provided the number of MUs was greater than one, the
dose delivered per MU could be expected to be within 2% of the expected value
for the first (slitless flight tube) accelerator. This applied at both 6 and 18 MV.
For the second accelerator, the corresponding figures were as follows. The dose
per MU was within 2% of the expected dose per MU when more than 3 MU
were delivered at 18 MV and when more than 1 MU was delivered at 6 MV. It
was found that changing from a slitted to slitless flight tube improved stability.
The detailed plots shown are very similar to those in the paper by Hansen et al
(1998) and Sharpe et al (2000b) presented curves for each energy and machine
configuration measured at three different time periods of within a year.
Beam flatness and beam symmetry was also measured and found, for
the slitless flight tube, to be within manufacturers specification after 0.6 s
corresponding to a delivery of 4 MU. For the slitted flight tube, the stability did
not arise until 1 s after beam turn on or approximately 7 MU. Measurements were
made in both the GT plane and in the perpendicular AB cross plane. Based
on these experiments, the authors were confident that calculation algorithms for
estimating relative output factors in a stable beam would apply to exposures as
short as 4 MU. The rest of the paper presented detailed discussions of output
factors as a function of field size and position and these were found to be very
stable and predictable provided the field sizes were properly collimated to the
expected field sizes.
Martens et al (2001a) have described the improvements consequent on
adding a fast-tuning magnetron (FasTraQ) into three Elekta travelling-wave linear
accelerators replacing the mechanical plunger with a solenoid-driven plunger in
the magnetron. It was found that start-up and inter-beam segment times were
typically reduced by between 24 s. Delivery efficiency for a variety of clinical
and quality-assurance programmes was improved by an average of about 30%.
82
Figure 3.18. The dose delivered per MU by SL20b measured as a function of the total
number of MUs delivered for (a) 6 MV and (b) 18 MV x-rays. This machine was initially
equipped with a slitted flight tube (SL20b-S) but was later upgraded to a non-slitted design
(SL20b-NS). The dose-per-monitor-unit acquired for the slitted flight tube (circles) is
compared with data acquired at two points in time following the upgrade (inverted triangles
and squares). (From Sharpe et al 2000b.)
The dose output was within 1% for a 2 MU beam segment. Dosimetric errors
were small even down to a 1 MU beam segment. Beam symmetries and flatnesses
were acceptable at all energies and dose rates showed no obvious degradation in
low-dose beam segments.
Budgell et al (2001a) have also described FasTraQ. In Manchester, it has
been fitted to a 7-year-old linear accelerator. It was found that the dose per MU
versus MU improved to better than 1% at 2 MU compared with figures of 4% for
the slitless flight tube and 5% for the slitted flight tube. For new accelerators, the
performance was not changed by the introduction of the fast-tuning magnetron in
terms of these parameters. The biggest effect was on treatment efficiency. The
start-up time fell from 8 to 3.6 s and the intersegment time fell to 1.7 s at 6 MV
and 2.2 s at 8 MV where the figures are both for a fluence rate of 400 MU min1 .
This was for stationary leaves so that leaf movement time was not part of the
measurement. Budgell et al (2001a) reported that the total time per prescription
fell by an average of 30.7%. The consequences are that one may now use more
83
segments to improve the conformality or, keeping the number as before, decrease
the overall treatment time.
Ezzell and Chungbin (2001) have studied, for a Varian linac, the effects of
CORVUS-generated segments with less than 1 MU per segment. Komanduri
(2002) has also commented that, as the number of segments for IMRT of the
breast increases, the number of MUs per segment will decrease possibly leading
to quality-assurance issues and concerns.
Rock et al (2001) and McClean and Rock (2001) have made measurements
of small (1 cm 1 cm to 3 cm 3 cm) fields defined using both a Varian 2300
CD and an Elekta SL18 accelerator and compared the data with that from the
HELAX treatment-planning system (Version 5.1). Data were measured using
an ion chamber, a diamond detector and film. The measured data included
leakage through the leaves, linearity of dose with MUs, x- and y-profiles, depth
dose in water and, output factors. Good agreement was found between the
treatment-planning system and measurements for depthdose data for fields as
small as 1 cm 1 cm for the Varian linac and 2 cm 2 cm for the Elekta
linac. However, comparison of the planning-system-generated profiles with the
measured profiles showed significant differences in both directions from both
machines at 2 cm 2 cm and 1 cm 1 cm for the Elekta linac. Output factors
also do not agree with measurements by as much as 20% for the Elekta linac
at 1 cm 1 cm and by 8% for the Varian linac at 1 cm 1 cm. Preliminary
investigations also showed that there may be problems with the use of both offaxis and rectangular fields. They concluded that caution should be used in the
MSF technique when small-field segments are incorporated.
Jones et al (2003) have pointed out that, for the beam components in
an IMRT plan which have a very small field area, the depthdose curve
will not necessarily follow traditional radiological scaling for tissue density
inhomogeneities. Using the EGS4 Monte Carlo dose-calculation package, graphs
were produced showing the depthdose curves as beams of either 6 or 24 MV
radiation passed through 3 cm of lung depending on the area of the beam element.
The situation is very complex.
Mitra et al (2001a) and Cheng and Das (2002) have investigated the
behaviour of two Siemens linear accelerators operating in IMRT mode at both
6 and 15 MV. It was found that the dose per MU was stable to 1% and linear
with increasing MU beyond 5 MU at 6 MV and beyond 10 MU at 15 MV. Below
these values, unacceptably large variations occurred. For example, at 15 MV
and 3 MU, the dose per MU was 5% above that required. At 6 MV and 1 MU
the dose per MU was 5% above that required. It was also found that profiles
were varying more than 5% below 10 MU at 15 MV. Above this, they also
showed that both accelerators displayed beam flatness and symmetry to better than
2%. This work was done by sequencing subfields using the Siemens LANTIS
and PRIMEVIEW system. They then concluded these characteristics make the
accelerators acceptable for performance of step-and-shoot delivery of IMRT. Saw
et al (2003) found different experimental results for a Siemens accelerator in
84
which the beam current was dephased from the microwave power whilst the field
shapes were adjusted. It was found, conversely, that the dose per MU was stable
to better than 2% above 2 MU.
Two consequences follow:
(i) IMRT delivery using this equipment should preferably have segments with
more than 10 MUs. When planning with HELAX TMS, the number of
segments should be kept low to ensure this condition.
(ii) Errors may occur in verification film dosimetry when MUs are divided down
in order to avoid saturating some types of film. In a typical case studied,
7/22 segments studied had less than 5 MU in this divided-down mode of
verification with film.
Bieda et al (2002) have shown for a Siemens PRIMUS accelerator that the
beam fails to reach full intensity during the delivery of 1 MU and a little more
than this is necessary for beam stability. This is relevant to IMRT with small
MU segments. Aspradakis et al (2002) have shown that the Siemens Primus
accelerator is ideally suited for step-and-shoot IMRT because its characteristics
in beam uniformity, energy and dose linearity are virtually independent of the
MU segment size. Li et al (2003) have made similar measurements to confirm
the suitability of a Siemens accelerator for delivering small-MU fields for breast
IMRT. Kulidzhanov et al (2003) also report on this problem.
Steinberg et al (2002) have improved the automatic frequency control
circuitry of a Siemens linac to create a beam which is stable within 55 ms instead
of the more usual 300 ms from onset of dose pulsing. This will create a situation
in which the delivery of a fraction of a single MU becomes reliably possible.
Bayouth and Morrill (2003c) also characterized small fields for IMRT.
Penumbra and leakage were remarkably consistent for the range of leaf positions
studied.
Cheng et al (2002) have compared the beam characteristics of Varian,
Elekta and Siemens accelerators, in particular noting differences in dark current
irradiation (see also Bassalow and Sidhu 2003).
In concluding this review, we should note that the beam performance
depends on the accelerator manufacturer, the specific accelerator, the segment
size and the version of the hardware. There are some inconsistencies between
reports from different groups on similar equipment.
3.5 Dynamic arc therapy

Dynamic arc therapy is the technique whereby, as the gantry rotates, through 2,
the field shape and intensity across it are changed by either using MLC movement
(figure 3.19) or some gravity-oriented device (GOD).
Pignoli et al (2000) have made measurements to compare with calculations
performed by an in-house-developed treatment planning system (TPS) for
Dynamic arc therapy
85
Figure 3.19. Illustrating the principle of the dynamic arc technique. As the gantry rotates,
the open MLC leaf aperture specified by a leaf pair defines the beams-eye view of the
target from that direction. Other leaf pairs operate similarly for the other transaxial planes.
This is little more than the principle of conformational radiotherapy of Takahashi shown in
figure 1.1.
the dynamic arc therapy technique delivered with the 3D-Line International
microMLC (see section 3.11.7). This technique constructs either three or five
arcs with the field shape of the microMLC tailored to the beams-eye view of the
target at each gantry location. Absolute isocentre dose was accurate to about 2%
and the calculated and measured isodoses matched to within 2 mm in transaxial
and sagittal planes.
Nakagawa et al (2000b) have developed a new accelerator capable of
dynamic conical conformal therapythey called this dynamic therapy. The idea
is that, instead of using couch rotation with respect to gantry rotation in a fixed
plane, the C-arm of the accelerator can itself rotate by up to 60 relative to
its default position, thus allowing non-coplanar conical conformal therapy. It
was shown that this improves treatment with respect to coplanar CFRT for large
tumours.
Boyer et al (2002a) have developed a sweeping window arc therapy (SWAT)
technique. In this, the leaves move along horizontal leaf tracks parallel to the axis
of a single rotation of the gantry.
86
Rivard et al (2002) have developed a doubly dMLC-IMRT technique in

which the gantry moves at the same time as the dynamic leaves.
Keane et al (2000) delivered total body irradiation using an arc and a gravityoriented compensator to achieve an intensity-modulation pattern. They also
changed the technique to use an MLC to obtain similar uniform dose distributions
through intensity-modulated-arc therapy (IMAT). By dynamically changing the
field sizes using the MLC as the gantry rotates, the equivalent to the custom
compensator can be produced.
Frencl et al (2001) have discussed how IMRT can be performed without
an MLC. The idea is to use a special treatment technique for head-and-neck
lesions. This comprises many static fields modified with a gravity-oriented block
for shielding the OAR, the spinal cord. Extra compensating filters are added to
improve the dose homogeneity in the PTV. Recently Danciu and Proimos (2003)
have compared various materials for making gravity-oriented absorbers.
Oh et al (2003) and Kim et al (2004c) proposed a new IMRT method based
on a scaled model of the PTV. The model hangs under gravity in a mercury box as
the gantry rotates. During this rotation, the beams-eye view of the PTV changes
conforming to the target and the fluence is modulated automatically. This is a
kind of variable compensator. Webb (2004a) has suggested an alternative to the
use of mercury and an inverse-planning method to calculate the optimum gravityoriented attenuator (figure 3.20).
3.6 Combining step-and-shoot and dynamic delivery for

dMLC
In general, IMRT is delivered with an accelerator equipped with an MLC in either
dynamic or step-and-shoot mode. Both of these could be considered embodiments
of the dMLC technique (the latter provided it comprised a large number of
segments). As pointed out by Martens et al (2001b), most departments choose to
work with one or other of these techniques to keep the investment costs sensible.
Alternatively, the equipment provided by a manufacturer may dictate the choice.
Dynamic delivery is more time efficient than step and shoot but the step-and-shoot
technique is less complex and has a lower cost of quality assurance. Martens et al
(2001b) have combined the best aspects of the two techniques in a hybrid method.
They call this interrupted dynamic sequences. It is not unlike the work proposed
by Beavis et al (1999, 2000b).
The way this works is as follows (figure 3.21). Consider, for example,
prostate IMRT delivered from three fixed gantry directions. A simultaneous boost
dose is delivered to the PTV minus the volume of overlap of the PTV in the rectum
using smaller segments and the whole prostate is also irradiated. This leads to the
combination of what the authors call a convex and a concave dose distribution and
is essentially an anatomy-based segmentation. When the segments are stacked
Combining step-and-shoot and dynamic delivery for dMLC
87
Source S
Rotation
through angles
0-2
S
Model of patient and

PTV
PTV model
'bath'
PTV
Figure 3.20. The delivery concept for a gravity-oriented device (GOD). The source is
represented at S. A section of the patient is shown in the lower part of the figure and a
demagnified version of this as a GOD between patient and source. The beam is collimated
to the beams-eye view of the PTV and modulated by the variable pathlength through the
absorber. Two attenuation coefficients bath and PTVmodel are determined by inverse
planning. This cannot effectively rival other methods of IMRT delivery.
together, this leads to a large dose gradient at the anterior rectal wall and typically
each of the beams may have a number of segments.
Considering just the segments required to realize just the intensity
modulation from one beam direction, the segments are classified into those that
belong to the part of the treatment delivering the convex dose distribution to the
whole prostate and the other part of the treatment which relates to the delivery of
the concave dose distribution. These two groups are called classes of segments.
The major step forward made by these authors is then that, for segments which
belong to the same class, dynamic transitions are allowed. The reason for this
is that for concave dose distributions corresponding to the segments that deliver
these, the intensity needed must continuously increase with decreasing distance
88
Figure 3.21. Segments of the three intensity-modulated beams with the MUs to be delivered for the interrupted dynamic sequences. Step-and-shoot
and dynamic transitions are represented by a plus sign and by an arrow respectively. (Reprinted from Martens et al 2001b with copyright permission
from Elsevier.)
IMATtechnical issues
89
from the rectal wall. Intuitively, the authors then argue that dynamic transitions
might actually be desirable because they blur the stepwise effect of segment edges
on intensity. However, a segment edge which is required to create a sharp in-field
intensity gradient should not be used as a dynamic sequence because blurring of
such an edge is undesirable and therefore step-and-shoot transitions are performed
between the segments for one class (those delivering the concave volume of high
dose) and the segment delivering the convex dose distribution. This is the basis
of the technique.
For two successive segments suitable for dynamic transition, an equal
number of MUs is transferred to the transition. The maximum number of such
MUs allowed for dynamic transition was 10. The accelerator was set to operate
at 100 MUs per minute and a complex detail in the algorithm relates to ensuring
that the dynamic transitions are selected in such a way as to minimize the overall
transition time. This actually means that some segments inside the concave class
cannot be delivered with dynamic transition but must be delivered with step-andshoot transition. Martens et al (2001b) then go on to make measurements using
a PTW-Freiburg LA48 linear ion chamber array in which the centre-to-centre
distance in the chamber is 8 mm and the whole array is shifted longitudinally
by 1 mm seven times to result in a measurement of a profile with a measurement
every 1 mm. They then make measurements for the same delivery using either
the step-and-shoot-only technique or the so called interrupted dynamic sequences
technique and they show that the two techniques are dosimetrically equivalent.
The new technique, however, is some 10% more time efficient. As part of this
work, they also make measurements of the relative dose output per MU as a
function of the MU delivered per segment and show that, provided that 3 MUs
or more are delivered, the deviations in dose output are less than 1% irrespective
of the dose rate. They also make measurements of field flatness under different
dose-rate conditions. In summary, they find the interrupted dynamic sequences
technique maintains the intended sharp gradients near the rectum but leads to no
major effects on the dose distribution in the regions of dynamic transition.
3.7 IMATtechnical issues

Yu and Li (2001) have developed a delivery optimization scheme with combined
rotational and fixed-gantry IMRT. This is because they do not believe that either
of these separately are optimal. An optimization algorithm was used to generate
beam intensities or beam angles, then one by one each beam was deleted from
the initial set and the intensities for the remaining beams were reoptimized. The
difference between the new and the baseline cost represented the importance of
the deleted beam. In this way, importance factors were evaluated for all beam
angles. The approach was then to deduce the optimal delivery technique such
that, where the importance factors peak at certain angles, fixed fields are used
at these angles, whereas when the importance factor as a function of angle is
90
Original IMBs planned at fixed gantry angles, in this example separated by 30

Beam profile at 00
Beam profile at 30
Angle for beam profile delivered via IMAT labelled to left of each subfield; multiple rotational arcs
0-30 rotation 1
3-60 rotation 1
0-30 rotation 2
0-30 rotation 3
3-60 rotation 2
3-60 rotation 3
Figure 3.22. This illustrates the principle of IMAT. Imagine the IMRT plan has been made
with IMBs at a series of fixed gantry angles (for illustration here they are at 3 separation
in angle. Imagine that each beam has three intensity levels. The logic of IMAT is that
there are three arcs and each of the individual fluence components is delivered spread out
over 3 of arc. The MLC leaves move as minimally as possible between arc segments. In
practice, the individual IMBs would be planned at a wider spacing S with more (say N)
fluence levels and there would be N arcs.
flat, rotational delivery will be chosen. The authors claimed this has a significant
clinical impact.
Yu et al (2004) have further developed the scheme for optimizing hybrid
IMRT. The first step in the scheme is to compute the angular cost function (ACF)
being a measure of how useful a particular beam orientation is with respect to
treating the target and sparing normal structures. The basic idea then is to treat
this function as the required delivery quanta at each of the possible angles and to
approximate it with delivery sectors. A delivery sector is a rectangle that can be
fit under the angular cost function curve. Long sectors are delivered by arcs and
short sectors are delivered with fixed fields. It was shown in a planning example
that the hybrid IMRT technique gives a better dose distribution than either fixed
field IMRT or IMAT.
IMAT, the multiple-arc IMRT technique invented by Yu (1995), usually
operates by first computing field modulations using an inverse-planning system
and subsequently and separately chopping up these modulations into arc patterns
for an MLC (figure 3.22). There are as many arcs as there are number of intensity
levels in the plan. A lay-scientist account of the work at the University of
Maryland on IMAT appeared in Wavelength (2000c) emphasizing the simplicity
of the technique, the requirement for quality assurance (QA) and that the method
is useful when it is desirable to spread out the unwanted normal-tissue integrated
dose. Wong et al (2002a) have presented an intriguing extension of the concept.
91
Their idea was to create field patterns for multiple-arc therapy in a single planning
step.
Their concept was to create a sequence of arcs with each arc having a
different planning goal. They also specified three levels of complexity for the
new technique. They rooted their notions in the concepts of arc therapy and (the
italian) bar-blocking therapy. All three strategies presented have one common
property, namely that the first arc creates MLC-shaped field outlines which
entirely span the PTV. The differences between the three strategies then centre
on how they cope with the organ-at-risk (OAR) sparing and in the balance of dose
to PTV and OAR.
Strategy 1 achieves the following:
(i) One arc uses the MLC to conform to the beams-eye view of the PTV in each
gantry orientation,
(ii) One additional arc for each OAR then uses the MLC to conform to the
beams-eye view of the PTV but shielding the OAR with the bank of leaves
that covers the least PTV view.
The total number of arcs is one for the PTV and one for each OAR.
Strategy 2 instead achieves the following:
(i) One arc uses the MLC to conform to the beams-eye view of the PTV in each
gantry orientation,
(ii) Two additional arcs for each OAR use the MLC to create an island shield.
The arcs use the MLC to conform to the PTV in each beams-eye view of
the arcs and shield the OARs in the beams-eye views with opposing banks
of leaves for the two arcs.
The total number of arcs is one for the PTV and two for each OAR.
Strategy 3 instead achieves the following:
(i) one PTV arc as before,
(ii) two additional arcs for each OAR as in Strategy 2 and
(iii) Two additional arcs per OAR that use the MLC to compensate dose nonuniformity in the PTV whilst shielding OARs. The extra arcs shield the
critical organ as before but irradiate the portion of the PTV only that wraps
around the critical organ.
The total number of arcs is one for the PTV and four for each OAR.
The techniques are operated in practice by weighting the contributions
differentially for each arc and selecting the weights that lead to an acceptable
plan. Wong et al (2002a) show examples of the MLC configurations and the
resulting plans. The indications are that the modified IMAT technique has great
flexibility and can provide a better therapeutic ratio compared with conventional
IMAT or conformal therapy.
Chen and Wong (2004) have developed an MLC leaf optimization scheme
for IMAT. This work started from the premise of two arcs, one conforming to
92
a PTV and the other to the PTV avoiding the OAR (called simplified IMAT
or SIMAT). Then, by dividing arcs up into 10 gantry angle increments, the
optimization variables become the MLC leaf positions with user-specified leafmovement constraints. Examples are shown of the improvement consequence on
this planning technique.
Shepard et al (2002a, 2003a) have developed a new inverse-planning
technique for IMAT that begins with a specification of the number of arcs and
incorporates the MLC and gantry constraints into the optimization. For this
reason, there is no loss consequent on the more traditional method of resequencing
inverse-planned modulated fields for IMAT. All constraints are observed on the
fly as the calculation proceeds.
Shepard et al (2003b) have put on the web a toolbox comprising data for
pencil-beam irradiation of a uniform cylinder together with tools for constructing
structures and beams-eye views of structures. The elemental beams are 1.0 cm
0.5 cm in size in a 20 cm 20 cm area with associated depthdose data stored in
a 125 125 31 cube. Data can be rotated and combined to field shapes. It is
argued that this permits the investigation of different optimization algorithms.
3.7.1 IMAT in clinical use
Yu et al (2000, 2002) have described their clinical implementation of IMAT.
This has been implemented at the University of Maryland School of Medicine in
Baltimore and by May 2001, 50 patients had completed their treatments with the
IMAT technique. The distribution of cases was 22 head-and-neck cases, 19 central
nervous system (CNS) cases and nine prostate cases. As far as planning for IMAT
was concerned, forward planning with a commercial system RENDERPLAN
3D was compared to inverse planning using CORVUS. Arcs were approximated
as multiple shaped fields spaced every 510 around the patient. The number
and ranges of the arcs were then chosen manually. It was found that between
two to five arcs were needed to achieve highly conformal distributions and that
the dose homogeneity to the PTV was high with IMAT. It was noted that the
CORVUS system has an inherent tendency to generate highly modulated fields
even when there are a very large number of fields, since there is nothing built
in to preclude this. This does not matter for the delivery of CORVUS plans
using the NOMOS MIMiC but inherently seems counterintuitive. In theory, the
more beams that are in use, the less the modulation needs to be required for
each beam and Yu et al argue that, for this reason, forward planning for IMAT
might be preferable to inverse planning for IMAT. They also compared the two
treatment-planning techniques described with a commercial forward planning of
3D conformal treatment plans not using intensity modulation. Not surprisingly,
they showed that the IMAT technique generates more conformal plans than the
conformal technique with just geometrical field shaping. They argued that,
provided the modulation does not change greatly from angle to angle, the field
shapes of the MLC also do not change much from angle to angle and so the
93
limiting feature is not the maximum leaf speed but the maximum gantry rotation
speed and the output of the accelerator. For 32 patients, comparisons were made
between measurements and predicted doses, using phantom measurements and
these were all less than 1% to the dose at isocentre. Care had to be taken to use a
dual-source model with consideration of leaf thickness and shape to calculate the
head-scatter factors for irregularly shaped fields. One feature that was emphasized
was that it was not a good idea to perform a full inverse plan and then to chop it up
for IMAT since the degree of modulation would generally lead to a large number
of field components and a large change of fields shape between gantry angles. In
summary, it was better to tailor the inverse planning to the IMAT technique.
Yu et al (2002) repeated the many advantages of IMAT. It can be
implemented on existing linacs equipped with an MLC. IMAT also takes only 10
15 min whereas five to seven gantry-fixed-angle IMRT takes 2030 min. Because
it does not collimate the beam into a slit, most of the target is in the field all
of the time, thus maintaining a high efficiency. No additional patient-transport
mechanisms are required to index the patient from slice to slice and therefore there
are no beam-abutment problems. They restated that the different MLC constraints
have to be built into the inverse-planning process. In summary, IMAT has been
successfully implemented using Elekta accelerators.
Li et al (2000b) have compared IMAT with fixed-gantry IMRT. It was
found that the use of seven fixed-gantry fields with ten intensity levels generated
worse results than IMAT for most of the cases studied even when only three
intensity levels were used in IMAT. The IMAT inverse planning was performed
by approximating the motion to fixed-gantry angles with increments of 5 . The
results should be evaluated in the context of studies by Rowbottom et al (2001).
Ma et al (2000d) have shown that IMRT of the prostate is possible using
IMAT with just two conformal arcs of 120 and 140. The IMAT plan was
computed using the RENDERPLAN treatment-planning system and the treatment
was compared with a seven-field IMRT treatment generated using the CORVUS
commercial inverse-planning system. Not surprisingly, IMAT and IMRT were
shown to be superior to conventional treatments but the IMAT treatment was
shown to give less rectal dose than the full IMRT treatment. IMAT led to 30% of
the rectal volume receiving over 50 Gy and 16% receiving over 58 Gy whereas
IMRT led to 33% receiving over 50 Gy and 14% receiving over 58 Gy. IMAT was
also simpler to deliver.
Cotrutz et al (2000) reported on a related form of IMAT. Wu et al (2001c)
have developed a new leaf-sequencing algorithm for IMAT based on the use of
graph theory. Iori et al (2003) have introduced IMAT in Reggio Emilia when
some dMLC IMRT techniques would be too complicated or lead to too great a
workload.
94

Original IMBs planned at fixed gantry angles, in this example separated by 30
Beam profile at 30
Beam profile at 00
Angle for beam profile delivered via AMAT labelled to left of each subfield
00
10
20
30
40
50
Figure 3.23. This illustrates the principle of AMAT. Imagine the IMRT plan has been made
with IMBs at a series of fixed gantry angles (for illustration here they are at 3 separation
in angle. Imagine that each beam has three intensity levels. The logic of AMAT is that the
three component intensities and field shapes are delivered at three separate angles spaced
by 1 with virtually no change in the resulting dose distribution. In practice, the individual
IMBs would be planned at a wider spacing S with more (say N) fluence levels and the
individual fluence components would be delivered at intervals of S/N 0 .
3.7.2 IMAT modified to aperture-modulated-arc therapy (AMAT)

Crooks et al (2003) have introduced a radically new method of IMRT delivery
known as aperture-modulated-arc therapy (AMAT). The acronym is deliberate to
draw an analogy with intensity-modulated arc therapy (IMAT). The logic of the
proposal is as follows (figure 3.23). Typically, IMRT using an MLC is delivered
at a set of discrete angular orientations with the MLC leaves taking up a series of
positions at each orientation. The new proposal is based on the logic that says that
the dose distribution from a radiation beam will change very little if that beam is
moved by a small angular displacement. The authors then propose that one might
imagine that, for an IMB with say ten component apertures defined by an MLC,
this could be represented as ten separate beams coming from angles very slightly
separated in asimuth with just one aperture at each of those angular orientations.
By extending this logic, a number of beams, each with a number of apertures, is
entirely reduced to a single aperture at a much larger set of angular orientations.
Unlike IMAT which uses several arcs, AMAT needs only one. The logic continues
that, by delivering the radiation in arc therapy mode with a constant output rate
from the accelerator, all that then has to be done is to modify the time for which a
New ideas related to the dMLC technique
95
particular aperture is exposed (and, in practice, this means modifying the angular
orientation over which a particular aperture is exposed) and then, by delivering
the set of apertures, so modified, using a constant output from the accelerator, the
overall delivered dose will be the same as that which would have been delivered
had a series of static components been delivered each from a number of fixed
beam orientations.
Crooks et al (2003) have written code to convert NOMOS CORVUS
distributions into such AMAT sequences. This was done for a number of planning
situations and shown, by delivering both conventional and AMAT deliveries to
film and mosfets, that the dose distributions are not vastly different. However,
they consider a close agreement to be of the order 710% and it is debatable
whether others would agree. Also there is very little change to the overall number
of MUs used nor to the time of delivery. However, the technique has a certain
elegance. Also it must be commented that the technique is limited by the ability
of the MLC leaves to change their shape rapidly during small arc rotations. All
these aspects are being investigated by the authors in further work.
3.8 New ideas related to the dMLC technique

MacKenzie and Robinson (2002) have presented a technique to deliver 24 fields
each modulated using the sliding-window dMLC technique and spaced at 15
intervals to simulate the effects of tomotherapy. This is essentially an extension
of the small-number-of-fields dMLC techniques to a larger number of fields. They
claim that this will overcome some of the drawbacks of tomotherapy, for example
the requirement to have specialized delivery equipment. An iterative filteredbackprojection algorithm was used to create IMRT fields which were interpreted
for the dMLC delivery technique. These were then delivered to a homogeneous
cylindrical phantom, initially constraining the problem to deliver a single axial
slice using a large number of projections. The delivery was performed with a
Varian Clinac 2300 CD equipped with a 52-leaf MLC and verification film was
placed along the transverse-axial plane coincident with the rotation plane of the
gantry inside a series of slices of the cylindrical phantom. By analysing the
experimental measurements in Fourier space and correlating with the predictions,
it was found that, for a variety of distributions, the measurements agreed with
the predictions to within about 3% which they argue is roughly the uncertainty
attached to the measurement technique. They then repeated the measurements
delivering the same modulation to a variety of slices by synchronizing all the
leaves.
96
Figure 3.24. This shows the basic building blocks of an Ellis compensator with some
arranged to make a 1D modulated profile by the compensation technique. The area of the
base of the blocks determines the spatial resolution and the height of the blocks determines
the intensity resolution.
3.9 Compensators and comparisons of compensator and

MLC-based IMRT
3.9.1 Do we need the MLC for IMRT?
A traditional way to create a modulated fluence over a shaped area has been
the use of a compensator. The first compensator was the so-called Ellis
compensator, a modern example of which is shown in figure 3.24. This
comprised LegoTM-like blocks of varying sizes and attenuating thicknesses.
Another way to make a compensator is to mill a metal block to the required shape.
Another way is to cut sheets of lead of different shapes and glue them together
to make a pattern (figure 3.25). A fourth way is to use a hot-wire cutting device
to cut a mould of the required shape out of tough Styrofoam and then to fill this
with either lead or tungsten balls or a hot liquid melt which cools to the required
shape.
Sherouse (2002) has written a perceptive letter extolling the virtues of the
metal physical compensator compared with the use of an MLC for IMRT. A few
sentences are worth quoting: There is a widely held misconception that IMRT
and MLCs are the same thing. IMRT without an MLC is like a fish without
a bicycle (think about it!). MLCs are not inherently an enabling technology
for IMRT but some form of computer optimization surely is. Inexplicably the
Compensators and comparisons
97
Figure 3.25. This shows how a compensator was constructed from thin lead sheets
glued together in order to modulate the intensity of a tangential beam for improving the
homogeneity of dose in breast radiation therapy. (Courtesy of the Breast Technology
Group, Royal Marsden NHS Foundation Trust.)
MLC enthusiasts seem to always conclude that the moving patient is the entire
problem, rather than question the wisdom of their choice to unnecessarily use
time-varying fluence patterns. He argues that the use of MLCs generate massive
re-education and training requirements and that the literature is filling rapidly with
reports on subtle and gross problems with the use of MLCs for IMRT. Finally, in
the US, IMRT with a compensator is billed less than with an MLC, thus driving
centres towards the latterthis is surely wrong. In short, Sherouse was, and still
is, a fan of the compensator. Chang et al (2004a) emphasized the advantages of
collimators in terms of static delivery with fine-scale spatial resolution and the
disadvantage as a limited dynamic range.
Welch and Harlow (2001) controversially have stated that a department with
just one MLC is likely to have excessive stress due to the uncertainties consequent
on downtime for this MLC. They argue that departments should have at least
two MLCs and, if the funding for only one is forthcoming, it might be better
to improve the block-cutting and compensating facilities instead of purchasing a
single MLC. This paper somewhat kicks against the trend for the increased use of
technology in radiation therapy.
3.9.2 Use of compensators for IMRT
It is well known that the use of IMRT improves the homogeneity of dose to the
breast particularly with large-breasted women (see section 5.7). Compensators
provide a particular delivery technique. The use of CT is largely avoided in the
98
UK because of the difficulty of access to a CT scanner for this large patient group.
Also, following work at Institute of Cancer Research/Royal Marsden Hospital,
compensators can be designed from portal imaging data. Compensators can
also be designed from measurements of the optical contour of the breast with
inferred lung positions. Wilks and Bliss (2002) have taken this a stage further and
proposed that a library of compensators can be reused for subsequent patients.
They planned 94 patients between June 1999 and May 2001. The first 28 led to the
production of 28 compensators used for treating those patients. The subsequent
66 patients were treated with the most suitable compensator selected from the
library.
Wilks and Bliss (2002) and Bliss and Wilks (2002) have produced
compensators by taking seven outlines using OSIRIS (QUADOSSandhurst).
A check film determined the lung thickness. Each compensator was built from
sheets of lead of thickness 0.5 mm. The technique for compensator selection
is as follows. A compensator could be individually fabricated. Alternatively, a
compensator could be manually selected from the library. More adventurously,
a compensator could be selected from the library by calculating the dose
distributions for using all field-size-appropriate compensators and selecting that
which produces the best dose-volume histogram (DVH) for the breast. The
compensator selected is then used to fully plan and evaluate the dose distribution
to the individual patient with inspection of isodose plots. The goal is not to
produce the best plan but an acceptable plan. On average, it was shown that
the use of the library compensator did not statistically degrade dose compared
with the use of an individually fabricated compensator.
Xu (2002) and Xu et al (2002) have invented a mechanism to generate
an automatic compensator for IMRT that is reusable and have characterized its
performance (figure 3.26). This is something genuinely new. A deformable
attenuating material comprising 50% tungsten powder, 35% silicone rubber and
15% paraffin was constructed. The linear attenuation coefficient of the material
was measured to be 0.41 cm1 and the maximum thickness of the compensator
was 10.2 cm which allows a transmission of 1.6% for 6 MV x-ray beams. A
set of 16 16 pistons were made to stamp down into this material to form a
compensator of previously calculated variable thickness. Each piston has a cross
section of 6.37 mm 6.37 mm which projects to 1 cm 1 cm at the isocentre.
The size of the pistons and location of the compensator were such as to create a
16 16 matrix of 256 cm2 bixels at the isocentre. The procedure was to stamp
the material out of the beam (but in the accelerator head) and then actuate the
compensator into the beam. Following use, the compensator was restamped to a
uniform thickness prior to the whole process repeating for another stage.
A limitation of the prototype was the non-divergent nature of the beam
leading to a penumbra increasing in width with distance from the central axis and
ranging between 5.510 mm. The thickness was determined for each bixel using
a simple logarithm of the ratio of input-to-output intensity and a division by the
linear attenuation coefficient of 0.41 cm1 at 6 MV. The percentage depthdose
99
Figure 3.26. The automatic physical compensator for IMRT mounted in the head of a
linear accelerator. The piston module stamps on the deformable attenuation material and
generates a physical compensator. This is then pushed into the line-of-sight of the beam
by a linear actuator. (From Xu 2002.)
curves determined using slabs of the material of thickness up to 4.4 cm showed

beam hardening at depth up to 4%.
Compensators were created and IMRT delivered to film. Calculations of
the delivered dose were made assuming a Gaussian spot and by single ray
tracing through the metal, scatter ignored. Comparisons of profiles between
measurements and calculations agreed to about 3%. Extracting data from each
bixel and plotting against thickness of compensator showed an independent
measure of agreed to 3% with that above.
This was preliminary work with a prototype and it may be imagined how
more detailed modelling of beam hardening and scatter could improve the fit.
The next stage is automation. The authors claim IMRT can be performed as easily
with the automatically reusable compensator as by the better-known MLC-based
methods. Xu et al (2003) report on a 32 by 32 model with improved spatial
resolution.
Yoda and Aoki (2003) have also developed a very neat technique for
producing a multiportal compensator system for IMRT delivery (figures 3.27
100
Figure 3.27. Developed compensator mount having a diameter of 55 cm and a height of

10.87 cm, which is placed on top of the linac head. The mount has six circular holes having
a diameter of 10 cm, so that six cylindrical compensator enclosures can be positioned. The
mount is rotated by a built-in stepping motor (which is seen behind the mount) and the
motor is controlled by a PC installed in an operating room next to the treatment room.
(From Yoda and Aoki 2003.)
3.29). Their system comprises a carousel which fits below the head of the
accelerator and contains six circular holes each of a diameter 10 cm, into which
preformed compensators can be inserted. The compensator carousel is rotated by
a motor and this is done automatically from outside the treatment room. The
estimated time for treated using five fields is just a couple of minutes for an
isocentre dose of 2 Gy with a 2 Gy min1 beam intensity. The rationale for this
development was that it is reported that currently available MLC-based IMRT
techniques have several clinical disadvantages and the multiportal compensator
system overcomes these. It has a shorter treatment time; it has less radiation
leakage; and there are no difficulties with MU calculations. Aoki and Yoda (2003)
have described apparatus using a 10 10 piston arrangement to stamp out the
compensator shape prior to pouring heavy alloy granules and vacuum forming
the compensator. Note this is distinct from the development of Xu (2002) which
uses piston stamping in the treatment room.
The individual compensators are made as follows. Firstly a set of 10
10 rods is forced into an appropriate shape to make the curved 2D surface
that will eventually become the final surface for the compensator. The rod
heights are determined from a 2D intensity map from a treatment-planning
system. The resolution is 0.5 mm. Each rod element has a square section of
101
Figure 3.28. Compensator system with Mitsubishi C-arm linac. Only five compensators
are installed so that the empty hole can be used for rotation angle calibration. (From Yoda
and Aoki 2003.)
5.97 mm 5.97 mm so that the spatial resolution of the intensity modulation is

10 mm 10 mm on the isocentre plane. After the rod heights are set-up, a preheated thermoplastic sheet is placed over the rod elements and vacuum forming is
performed. Then heavy alloy granules with a density of 10.1 g cm3 are poured
into the vacuum-formed shape. These are tightly packed in by screwing down a
pressure plate and then plastic tape is applied to hold all the granules in place.
Once prepared, all five compensators are placed in the carousel and the whole
process takes about 20 min. The authors have delivered an IMB using 6 MV xrays from a Mitsubishi C-arm linac to a film and shown that the required dose
distribution roughly matches that measured. The linear attenuation coefficient
of the material was 0.0515 mm1 . This represents a very nice technique for
compensator-based IMRT.
McCurdy et al (2002) have developed an automated compensatorexchanging device which enables compensators to be used for IMRT, the new
compensator being selected at the same time as the gantry angle changes.
Warrington et al (2002) have developed a carousel suitable for housing
up to four conformal lead-alloy blocks of maximum field dimension 10 cm for
selected conformal or IMRT prescriptions delivered with an MDS Nordion Elite
80 tele60 Co machine (figure 3.30). This has a 2-cm-diameter source and an 80 cm
source-to-isocentre distance. Studies showed that the system could deliver beams
with a satisfactory accuracy of between 1 and 2 mm at isocentre.
Peltola (2003) has designed compensators using tin-granulate and lowmelting-point heavy metal alloy, using the CADPLAN HELIOS system and
102
Figure 3.29.
View of the 10 10 rod elements after adjusting to rod height
according to the desired intensity profile. Each rod element has a square cross section
of 5.97 mm 5.97 mm so that the spatial resolution of the intensity modulation is
10 mm 10 mm on the isocentre plane. The rod has slightly cutout corners so that vacuum
forming can be performed by drawing air downward through the cutouts. (From Yoda and
Aoki 2003.)
103
(a)
(b)
Figure 3.30. (a) The carousel equipment for delivery of radiation of different geometrical
shape and fluence using a 60 Co machine. In the form shown here, only the geometrical
shape can be varied but the inserts could instead be compensators. (b) The components
of the apparatus shown dismantled. The two plates on the right are a jig for pouring
the cerrobend block with the result shown to the left. (Courtesy of J Warrington, Royal
Marsden NHS Foundation Trust.)
104
Figure 3.31. Delivery of multiple IMRT treatment fields with one multi-field modulator.
The gantry isocentre (small filled circle) is placed outside the target volume (paranasal
sinus cancer). The gantry and collimator rotate to their first positions as the jaws collimate
the first field. Additional collimation can be provided by the MLC leaves, blocks or full
thickness of the modulator. The beam is switched on and the first field is treated. Then the
gantry and collimator rotate to their second positions and so on to complete the treatment.
(Reprinted from ODaniel et al 2004 with copyright permission from Elsevier.)
showed that, with suitable correction factors, beam-hardening effects can be

overcome.
Gurgoze and Rogers (2002) have described what they call MLC replicators
(MLCRs). By this they mean compensators. They show that the multileaf
replicator plan and the MLC plan agree to within 3% and that the multileaf
replicator can deliver IMRT in 7 min compared to 17 min for the MLC.
Papatheodorou et al (2000b) have developed a virtual compensation
technique using the dMLC technique. The primary-scatter separation dose model
of the treatment-planning system was extended to account for the intensity
modulation generated by the dMLC.
ODaniel et al (2004) have also constructed a rotating multifield
compensator. For each field, the compensator rotates to have exposed a different
pattern of metal thickness. It was shown that use of this device gave conformal
distributions for treating paranasal sinus cancer that were as good as the use of an
MLC (figure 3.31).
105
3.9.3 Comparison of compensator and MLC-based IMRT

Chang et al (2000b, 2002) have compared the dosimetric outcome of performing
IMRT with either a compensator or the MSF-MLC technique. They considered
two tumour situations and used the inverse-planning system PLUNC to create
modulation maps, having first specified the number of fields. They then, for
the MSF-MLC technique, stratified these maps into a discrete number of fluence
intervals. They varied this number and studied the effects. The compensators
were constructed from tin-granulate poured into moulds whilst the stratified
skyscraper-like maps were fed to the Siemens IMFAST sequencer to determine
the field components. Dose distributions were computed and compared for the
two techniques.
They introduced an interesting index, the Uniform Dose Volume (UDV),
being the volume of the PTV raised to 95% or more dose. They showed that this
UDV index was largest (of course) for the unstratified and spatially continuous
compensator technique. The UDV for the MSF-MLC method increased as the
number of modulation intervals increased. From the behaviour of this index,
they determined that five levels were sufficient for the cases studied. They also
showed that the number of field segments generated by IMFAST increased in
proportion to the number of segmentation levels and was typically between one
and two segments per level per field. Interestingly, it was found that the biggest
differences between the two classes of IMRT technique were due to the finite
spatial resolution of the MSF-MLC technique and not to the number of fluence
levels selected. This was particularly so for the protection of the OARs which
they quantified by studying the integral DVHs. As well as using IMFAST, they
computed segments independently using PLUNC and obtained the same number
of segments but with slightly modified MUs due to the two-Gaussian model used
by PLUNC for head-scatter calculations compared with a one-component model
in IMFAST.
In conclusion, Chang et al (2000b) found that the compensator could be
regarded as a gold-standard IMRT treatment against which the degradation
introduced by the MSF-MLC techniques could be measured. Chang et al (2003)
continued this theme pointing out the superior fluence and spatial resolutions of
compensators for 340 patients. However, they also point out the limitations in the
dynamic range of fluence using some tin-based compensators.
West and Jones (2000) have also compared the dose distributions delivered
from physical compensators and multileaf compensation with the MSF technique.
Not surprisingly, it was found that the MSF technique degraded as the number of
fields applied was reduced.
Grigereit et al (2000) have developed compensators for IMRT and showed
that delivered dose distributions were within 1% of measured dose distributions.
Jursinic et al (2002) have delivered IMRT using five fields with
compensators milled from brass with a spatial resolution of 2 mm and shown
that the delivered IMRT plans are better than those using MLC modulation.
106
Kermode and Lawrence (2001) have developed a practical system to use the
HELAX treatment-planning system to design simple 2D compensators to account
for changes in patient contour or specific OAR shielding. The method was verified
by comparing the predicted and measured beam profiles. Kermode et al (2001a)
gave data on the improved dose distributions in the head-and-neck regions due to
the use of compensators.
Schefter et al (2002) have shown that dose distributions created with a
compensator and with MSFs were very similar when treating cervix cancer using
five fields. Increasing the number of fields did not improve the treatment. They
argued that the increased simplicity of using compensators gives this technique
an advantage.
3.10 Optimum width of leaves for an MLC

Oelfke (2001) has described the features of MLCs that are required for IMRT.
Bortfeld et al (2000) have applied sampling theory to the problem of determining
the optimum leaf width of an MLC for radiation therapy. The work of Bortfeld et
al (2000) is predicated on ensuring that the sampling is optimized in relation to
the physical width of the point-spread function of the radiation. They showed that
this can be derived from measurements of the 2080% penumbra at the edge of
a step-function fluence and it turns out that the standard deviation is equal to the
width between the 20% and 80% dose contours divided by 1.7. From this, it could
be concluded that, for a 6 MV beam in soft tissue, the optimum sampling distance
is 1.52 mm. The most obvious next step is to make the physical size of the
MLC leaves equal to this sampling width and examples shown later in the paper
for both IMRT and for the delivery of fixed static fields show that, when this is
done, the effects of the collimator size are not observed in the dose distributions.
They pointed out that it is actually an adverse step to try to reduce the width of
the collimator below this as this would, due to sampling considerations, lead to
aliasing in the distribution.
They drew a distinction between the sampling distance and the leaf width
and showed that, provided the sampling distance is maintained as before, the leaf
width can double and the same results can be achieved by a double delivery
of radiation with the collimator moved by half a leaf width between the two
components of the irradiation. This is the so-called shift technique. Using an
IMRT example, they showed that this leads to a dose distribution which is much
the same as that delivered by a single delivery with leaves of the original small
2 mm leaf width. However, examples with the leaves considerably increased in
size demonstrate that the dose distribution deteriorates. Finally, they commented
that sampling theory can be used to find adequate beam element sizes and voxel
sizes for dose calculation and inverse-planning algorithms.
Nill et al (2002) have compared IMRT dose distributions delivered with
MLCs of different leaf widths. They studied two particular so-called internal
MicroMLCs for IMRT
107
MLCs and two particular external MLCs and the overall conclusion was that the
smaller the leaf size the better the homogeniety of dose to the target. However,
reducing the leaf width to 2.75 mm, whilst it resulted in further enhancement of
the target coverage, did not change the dose to OARs. They are still questioning
whether this is of any clinical significance.
Wang et al (2003b) have compared the use of two MLCs of leaf widths 4
and 10 mm to sequence a CORVUS plan and showed improved OAR protection
consequent on the use of the former.
Drzymala et al (2001) have studied the geometric aspects of IMRT planning.
They particularly looked at the variation of dose conformality with changing the
MLC leaf width, collimator angle and couch rotation. They concluded that it is
always best to use narrower leaf widths where possible and that when treating
with multiple coplanar beams it is desirable to avoid parallel-opposed beams.
3.11 MicroMLCs for IMRT

There are techniques for using a conventional MLC with leaf widths projecting to
1 cm at isocentre with shifts and multiple deliveries. These simulate the use of a
microMLC and are here referred to as virtual microMLCs. Planning the required
modulations has been described by Hou et al (2004) who also showed improved
smoothing properties of the developed IMBs (see section 6.5.9.) Reports of using
three manufacturers MLCs follow and after that specific microMLCs will be
reviewed. There is no clear distinction between a microMLC and a miniMLC so
the former name will be used here although some companies prefer the latter.
3.11.1 Siemens (virtual) microMLC
Siochi (2000a) has developed an algorithm for minimizing the number of
segments and junctions in virtual microMLC IMRT. This is the technique in which
a 10 mm 5 mm bixel intensity map is combined with that of a 5 mm 10 mm
bixel intensity map delivered at orthogonal collimator settings. The aim was to
minimize the number of segments and junctions for dosimetric purposes.
Siochi (2000c) has presented in great detail the way in which a conventional
MLC projecting to a leaf size of 10 mm at isocentre can be used to deliver
an intensity modulation over a grid of 5 mm 5 mm. The technique relies
on delivering two intensity maps at collimator settings that are 90 apart. The
superposition of these two component maps in which the leaves can be discretized
to 5 mm along their length in each direction will yield the required outcome. The
solution is non-unique and most of the study concerned the specific choices that
need to be made to do an efficient decomposition.
Mitra et al (2001b) have studied the Siemens HD270 MLC which is a
virtual microMLC that smooths the defined field edges by a series of coordinated
incremental couch movements. The movements can be in increments of 5, 3 or
2 mm over a 27 27 cm treatment field size corresponding to 1, 2 or 4 shifts.
108
The study showed that, from looking at dose at depth, negligible improvement in
target coverage resulted from increasing the leaf resolution to better than 3 mm.
Cheng et al (2000) have studied the dosimetry of a virtual mini multileaf and
its clinical application. The virtual mini multileaf in question was the Siemens
HD270 which combines the use of a large-leaf MLC with table translation
perpendicular to the leaf plane of either 2, 3 or 5 mm. Film dosimetry showed
that the 8020% penumbra width for a 45 block and a circular block were
reduced using the HD270 compared with the use of the wide-leaf MLC. At 2 mm
resolution, the scalloped isodoses virtually disappeared, although the penumbra
was still somewhat larger than for a cerrobend block. The HD270 was also shown
to improve the dose-volume histogram of the rectum compared with the use of a
broad-leafed MLC.
Twyman and Thomas (2002) have also described the Siemens HD 270 which
uses three couch shifts in conjunction with the use of a 1-cm-wide-leaf MLC to
simulate the action of an MLC with 2-mm-wide leaves.
3.11.2 Varian (virtual) MLC
Greer et al (2003) demonstrated experimentally that an improvement in fluence
resolution perpendicular to the direction of leaf movement could be obtained by
combining the fluence patterns from two Varian MLC deliveries shifted by half a
leaf width with respect to each other. This reduces the leaf sampling distance and
improves dose resolution. They made film measurements for several test patterns:
a bar pattern, a clinical field from a parotid plan and a doughnut of fluence. The
results were in line with the conclusions of Bortfeld et al (2000). Two different
sampling schemes were investigated.
3.11.3 Elekta (virtual) microMLC and microMLC
A simplified account of the work at the Royal Free Hospital using an Elekta MLC
appeared in Wavelength (2000b) emphasizing that pseudo high-resoution MLCs
can be fabricated by the isocentre shift method. This has been in use since 1994
for nasopharynx cases and for all routine prostate cases. A side effect of the
phased-field technique (PFT) is the reduction of interleaf leakage by about 50%.
The new Elekta MLC is a high-resolution MLC that will be IMRT
compatible (figure 3.32). It will be optically verifiable. It has a 4 mm leaf pitch
and is fully integrated into the treatment of the Elekta accelerator head. It can
generate a 16 cm 22 cm field. It has no back-up jaws because the leaves are
set in a solid millstone. It will allow interdigitation. Mosleh-Shirazi (2002) has
characterized the beam properties of this prototype Elekta high-resolution MLC.
The along-the-leaf and perpendicular-to-the-leaf 2080% penumbral widths were
4 and 3 mm respectively. The MLC replaces the lower jaws and the whole field
size is modulatable for IMRT.
MicroMLCs for IMRT
109
Figure 3.32. The Elekta Beam Modulator MLC. (Courtesy of Neil Harvey, Elekta
Oncology Systems, UK.)
3.11.4 Radionics microMLC

Hoban et al (2001) have used a Radionics microMLC for delivery of IMRT
(figure 3.33). By March 2001, five patients had been treated. The microMLC was
attached to a Siemens Primus linear accelerator. The Radionics X-plan treatmentplanning system with the integrated planning module from KONRAD was used
for inverse planning. Field sequences and intensity maps were then delivered
to a phantom for comparison of planned and delivered dose distributions. TLD
measurements were also made. Hoban (2000) has developed a model for
dosimetry for the micro-dMLC technique. This uses a Gaussian source, headscatter function and a pencil-beam kernel convolved with the opening density map
with the addition of radiation leakage. This calculation matches measurements
quite well. Hoban (2002) has used a microMLC to deliver a simultaneous IMRT
boost in the prostate.
110
Figure 3.33. The Radionics microMLC. (Courtesy of Vicky Aerts at the Radionics
Company.)
Parker et al (2000) compared dose distributions from geometric

conformation with IMRT for intracranial lesions. They also made use of the
Radionics microMLC and found that IMRT was required when sparing nearby
normal tissue is a high priority. Otherwise geometric conformal radiosurgery was
sufficient.
Solberg et al (2003) have used a microMLC to treat prostate cancer by IMRT.
Wang et al (2003b, c) showed that the rectal-and-bladder sparing was improved by
using the Radionics microMLC compared with the use of a conventional Siemens
MLC. Wang et al (2003bd) have also demonstrated the dosimetric advantage of
the microMLC in treating prostate cancer with IMRT.
3.11.5 BrainLAB microMLC
Barker et al (2001) have compared the dosimetric characteristics of a standard
Varian 52-leaf MLC and a BrainLAB M3 microMLC for square, rectangular
and irregular fields at 6 MV using a Varian Clinac 2300 CD linear accelerator
(figure 3.34). Conventional percentage-depthdose data were unchanged for most
situations. The microMLC provided a sharper penumbra. The relative dose
factors for a given MLC or microMLC field size depend on the jaw settings as
well as on the tertiary field size.
Benedict et al (2001) compared IMRT delivered with the BrainLAB M3
microMLC with stereotatic radiosurgery performed using 15 fixed non-coplanar
fields and arc-based stereotatic radiosurgery with 45 arcs per isocentre. The
MicroMLCs for IMRT
111
Figure 3.34. The BrainLAB microMLC. (Courtesy of the BrainLAB Company.)
IMRT technique for three of the four patients studied significantly reduced the
ratio of treatment volume to target volume.
Yenice et al (2001) have made plans for static and IMB delivery to be
delivered with a BrainLAB M3 microMLC. It was found that the use of the
microMLC consistently improved upon static CFRT.
Agazaryan and Solberg (2003) have written a sequencer for both static MLC
and dMLC versions of IMRT for the BrainLAB MLC on a Novalis accelerator.
They incorporated leaf transmission and a parameter to vary the number of leaf
patterns.
3.11.6 DKFZ-originating microMLCs
3.11.6.1 The MRC systems GMBH/Siemens Moduleaf
The Deutches Krebsforschungszentrum (DKFZ), Heidelberg, have designed and
constructed a new MLC for both stereotactic radiotherapy and IMRT. It has
leaves which project to 2.5 mm at the isocentre and so is intermediate between
the large-scale MLCs and the microMLCs. It is marketed by MRC Systems
GMBH/Siemens under the name Moduleaf (figure 3.35). There are 40 leaf pairs.
Each leaf is 7 cm high and the maximum field size at isocentre is 12 cm by 10 cm.
112
The leaves are driven by motors which are arranged in banks to drive alternate
leaves from the top or the bottom of the carriage. The leaves carry positionsensitive potentiometers. They can overtravel by 5.5 cm (see table 3.1). Hartmann
and Fo hlisch (2002) reported measurements to characterize the performance.
The leaves are focused in the y -direction by the convergent shape of the
leaf cross section although they are not precisely focused to the source in order
to remove the need for TG arrangements. The focusing along the leaf (x )
is quasifocusing through a three-sided leaf-end pattern. This leads to a mean
penumbra of 3.2 mm for the x -profiles and 2.9 mm in the y -profiles. The effective
penumbra for a 45 edge is 3.5 mm. The actual leaf positioning is accurate to
0.1 mm. The mean transmission is 1.3%. The small penumbra and accurate
positioning implied suitability for stereotactic radiotherapy and for IMRT.
3.11.6.2 New DKFZ microMLC
Pastyr et al (2001) have designed a new prototype MLC at DKFZ, Heidelberg.
This has physical leaf thickness below 3 mm allowing a leaf resolution of
below 5 mm at isocentre. It also has linearly-guided double-focusing leaves
to ensure that the edges are always directed towards the focus and yielding a
very small position-independent penumbra at all positions of the leaves (Forced
Edge Control). There is a simple high-precision linear leaf-guidance system.
The MLC was designed for IMRT. It does this by attaching a rotating cog to
two racks alongside each other but with slightly different teeth numbers per
unit distance. One rack drives the leaf. The other drives a pivoting leaf edge
(figure 3.36). Hartmann et al (2002b) further described the mid-sized MLC
developed at Heidelberg which has a mechanism for automatically focusing the
leaf in the direction of leaf motion irrespective of leaf position. The MLC also
has extremely flat potentiometers which are used as pairs embedded in each leaf
to control and additionally check the leaf position with high accuracy. Tu cking et
al (2003) at Heidelberg have shown that the use of the microMLC impoves dose
conformality.
3.11.7 3DLine microMLC
Mapelli et al (2001) have described a new microMLC from the 3DLine company
(figure 3.37). This has a redesigned leaf shape to reduce the TG effect. Leaf
transmission is less than 0.4%; leakage between leaves is less than 0.5%;
penumbra is less than 4 mm and the accuracy and dose delivery at the prescription
point in dynamic therapy is better than 2%.
3.11.8 Comparison and use of microMLCs
Hover et al (2000) have compared three different microMLCs for delivering
IMRT. Their results show that not all the collimators could be used for IMRT.
MicroMLCs for IMRT
113
(a)
(b)
Figure 3.35. The DKFZ/ MRC-Systems/Siemens Moduleaf MLC (a) with its covers off,
(b) with its covers on. (Courtesy of Professor Wolfgang Schlegel; Heidelberg.)
114
(a)
(b)
Figure 3.36. (a) Showing how one cog can operate simultaneously on two linear gear
trains to drive one train a little faster than the other and so make the two move relative
to each other. This has the effect of (b) turning the end of the leaf so its face is always
tangential to the divergent radiation from the source. (Courtesy of Professor Wolfgang
Schlegel; Heidelberg.)
MicroMLCs for IMRT
115
Figure 3.37. 3D Line microMLC. (Courtesy of the 3DLine Company.)
Leakage measurements showed a wide variation up to a factor of four between the

MLCs. Hover et al (2001) have further reported on the dosimetric characteristics
of five add-on MLC systems with small leaf widths. These were from Radionics,
BrainLAB, MRC, MRC/DKFZ and DKFZ. Measurements showed that IMRT can
best be done with double-focused systems whereas stereotactic treatment may be
done with cheaper single-focused or even parallel-sided MLCs.
Gibon et al (2000) have presented a method to optimize microMLC
irradiation for IMRT. The positions of the fields were optimized using a genetic
algorithm and the leaf position and the weighting factors of fields were optimized
by a simulated annealing method. The results showed a conformal improvement
by using IMRT and/or by increasing the number of fields. However, in many
cases, an optimized three-field microMLC technique with IMRT obtained results
which were almost equivalent to those obtained with non-optimized five-field
techniques. The results depended on the shape of the target.
Yang et al (2002) have shown that the use of a microMLC implemented
as a tertiary device can improve the high-precision requirements of stereotactic
radiotherapy.
3.11.9 Multi-level MLC
Topolnjak et al (2002) have described a new collimation system being developed
in Utrecht. One aim is to produce a 40 cm 40 cm field with the same resolution
as could be obtained with a microMLC. This will have six banks of a multileaf
system in three levels with each paired bank at 60 to the paired bank above
116
and below it. The banks of leaves are of conventional width but this allows the
simulation to mimic the microMLC. Because the tongues and grooves of one
particular bank are always overshadowed by leaves from a bank above or below,
it might be possible to abolish the TG mechanism altogether, thus simplifying
the design. Interleaf transmission is always blocked by at least one leaf from
another bank. Also there would be no need for collimator rotation. The group
have performed a computer design study. Topolnjak et al (2003a) has reported on
Monte Carlo modelling of the six-bank multileaf system for IMRT and Topolnjak
et al (2003b, 2004a, b) have described a recursive sequencer for this three-level
MLC which is still at the design stage.
Potter et al (2003a) have alternatively described a two-level microMLC
(Alanya Enterprises Corp, Paris) to improve spatial resolution.
3.12 Increasing the spatial resolution of a conventional MLC

Williams and Cooper (2000) have presented a study of the effect of the use of
a tertiary grid slit collimator applied to improving the spatial resolution of a
conventional MLC. The collimator comprises a plate with slits of width spaced
at intervals of W where W is the width of a conventional MLC leaf. W/
sequential irradiations take place with the collimator successively indexed by
for each irradiation. The prototype tertiary collimator was made of brass of
thickness 8 cm with parallel-walled slits focused to the radiation source. It was
mounted in the blocking tray of an Elekta SL25 accelerator.
The advantages of this device over a micro or miniMLC were said to be:
(i) high spatial resolution can be obtained over a field as wide as can be
collimated by the conventional MLC;
(ii) no seriously increased penumbra;
(iii) no requirement for the pseudo microMLC methods based on the use of a
conventional-width MLC.
Limitations included:
(i) the requirement for extremely precise incremental relocation of the tertiary
collimator;
(ii) a small increased irradiation time;
(iii) degradations due to transmission through the shielded parts of the collimator;
and
(iv) a potential reduction in uniformity due to multiple-subfield superposition.
Test irradiations at 6 MV were made to characterize the elemental radiation
profile of a single irradiation. Measurements were made at a depth of dmax and at
a depth of 10 cm for slits of with 2.5 mm and 5 mm. (For a conventional MLC
with W = 10 mm use of these tertiary collimators to provide a shaped field would
therefore require four and two sequential irradiations respectively.) The following
were observed:
117
(i) The variability in peak height of irradiation within the slits was greater for
the 2.5 mm collimator possibly due to difficulty of accurately machining the
slits;
(ii) the lowest intensity reached in the shielded regions was higher for the 5 mm
collimator.
The effect of combining multiple irradiations with sequential advances
of the collimator was determined by two methods. Firstly, the data for a
single irradiation were reused computationally to calculate the effect. Secondly,
measurements were made. Both showed that the variation in intensity was within
3% for the 2.5 mm collimator and somewhat worse for the 5 mm collimator.
When shaping a sinewave-shaped field of varying amplitude and periodicity,
the expected field-shaping benefits of the tertiary collimator were observed with
little degredation in penumbra (figure 3.38).
The effect of misregistering subfields was shown to be about 18% mm1 .
Cooper et al (2001a) have investigated whether radiation dosimetry parameters
change when the grid slit collimator is in use, concluding that no significant
differences were identified. The grid factor was found to be 2.95.
Cooper et al (2001b, 2002) have further discussed the grid slit collimator.
The prototype collimator had a leaf pitch of 10 mm equal to the width
of a conventional multileaf. It was found that artefacts could arise due to
mispositioning of the slit collimator with respect to the edge of a leaf when this
was defining part of a field. These artefacts were of two types. The first arose
because of the precision with which the grid could be manufactured, the slit width
used and the accuracy of the indexing. A second more important artefact was due
to the partial overlap of a slit and the junction between two leaves (which set the
field width for that slit). In this situation, the two leaves seen essentially define
two different widths leading to narrow fingers of overdose just outside the field or
underdose just within the field.
These artefacts were overcome by constructing a second prototype with 5mm-wide slits on a pitch of 15 mm in such a way that, if one slit were located at
the centre of one leaf, the next would straddle the junction of the next two. With
both adjacent MLC leaves being used and set to the same position to define the
slit length the edges of any of the slits were never required to be adjacent to a leaf
edge and this removed the artefacts.
Cooper et al (2001c) have studied the issue of the best way to index this
tertiary collimator to ensure that abutting irradiations correctly junction with each
other (figure 3.39). They considered two methods. In the first, the tertiary
collimator remained stationary and the patient couch was incremented by the pitch
of the open slits of the collimator (consider a tertiary collimator with mark-tospace ratio of unity, made of brass 8 cm thick and with a slit width projecting to
5 mm at isocentre). The second indexing technique was to arrange for the tertiary
collimator to move on the arc of a circle focused back to the source location.
Again, to achieve an increment of 5 mm at isocentre the tertiary collimator
118
Figure 3.38. The three images show the isodose patterns (20%, 50% and 80%) for a
4-cm-wavelength high-dose area created under three separate delivery conditions: (a)
created with the conventional MLC, (b) created with two grid irradiations of size 5 mm
and (c) created with four grid irradiations of size 2.5 mm. Notice that the isodose lines are
more faithfully reproducing the desired pattern of high dose using the tertiary collimator.
(From Williams and Cooper 2000.)
119
Figure 3.39. Photograph displaying the current form of the tertiary collimator from the
Christie Hospital, Manchester and its method of attachment to an Elekta SL accelerator.
(From Cooper et al 2002.)
needs to increment by a rotation of 0.005 rad or 0.29. Clearly the first method
can only lead to a good match at the isocentre because the open slits diverge.
Hence, a variation in dose at some other depth would be expected. Cooper et
al (2001c) made measurements of the radiation profile along the direction of
potential collimator movement, at both isocentre and at several depths above
and below. They then digitized the film measurements and computationally
added the appropriately shifted profiles. Whilst, for pure translation, the dose
inhomogeneity was only about 8% at isocentre, it rose to some 20% at planes
3 mm above and below. However, when the profiles were computer summed
using the rotation method of indexing the dose inhomogeneity remained at about
9% for all planes including the isocentre. This method of computational addition
was used to simulate perfectly accurate additions. Then the whole experiment was
performed by actually double-irradiating film with the tertiary collimator rotating
in its cradle. The dose inhomogeneity rose to about 15% but was constant for all
planes.
The originators of this concept had hoped to be able to produce a device
which generalized the use of an MLC with conventional-width leaves and
interest the appropriate manufacturer. The use of mini- and microMLCs
somewhat overtook the concept but at a considerable price (Williams, private
communication).
120
3.13 Verification of MLC-delivered IMRT

With the rapid proliferation of systems for delivering IMRT in a clinical setting,
using an MLC, the topic of verification has become very important. There has
been a general trend to widen the scope of techniques from one-off to those
which can be usefully applied in many clinics. Low (2001) and Budgell (2002)
have summarized verification methods for IMRT. In this section, verification of
specific clinical IMRT deliveries is described. The subject of routine verification
of the performance of an MLC is described in section 3.14. This is often referred
to as QA rather than verification. Boyer et al (2002b) also identified aspects of
QA of IMRT that are related to the delivery equipment and those aspects which
are patient-specific.
Ramsey et al (2003) have debated (in a Medical Physics Point and
Counterpoint) the wisdom or otherwise of individually verifying each IMRT
treatment plan. An argument on one side is that measurement only validates
a specific treatment plan and does not address issues of planning assumptions,
movement etc. An argument on the other side is that time freed from such
individual verification could be better spent on system QA to the benefit of all
IMRT patients.
3.13.1 Electronic-portal-imager-based IMRT verification
Partridge et al (2000b) have used electronic portal imaging (EPI) to verify both
the leaf tracking and also integrated intensity measurements. Modifications have
been made to a commercially available system to give accurate knowledge of
the cumulative delivered dose in every frame of a sequence for leaf-tracking
measurements and secondly to reduce optical crosstalk in the camera-based
systems. Portal imaging was demonstrated to be a useful tool for the verification
of the dMLC delivery technique.
Partridge et al (2000a) have made a comparison of three particular
implementations of the use of electronic portal imaging for leaf-position
verification during the dMLC IMRT delivery technique. The three systems
investigated were in three different centres namely those at the Royal Marsden
NHS Foundation Trust, the Christie Hospital NHS Trust and The Netherlands
Cancer Institute, Amsterdam. The goal of the work was to characterize the
performance of the leaf-and-jaw motion and to compare measurements with
prescription predictions. The first is an inhouse development; systems at the
second and third centres are commercially available. This investigation is
different from the investigation of total integrated fluence during the dMLC
technique since this latter does not provide information about what caused any
potential error. The measurements described by Partridge et al (2000a) give
information about the actual trajectories of each leaf and jaw. All measurements
were made using the Elekta dMLC delivery technique as existed at that time. Ten
different prescriptions were used. In any one prescription, either a jaw or a set
121
of leaves moved. In the first eight prescriptions, the jaws and leaves moved at
constant speed and in the ninth and tenth prescriptions the jaws and leaves moved
in a sequence: slow, fast, slow. This was to test whether the measured positions
of the leaves and jaws stayed within tolerance, noting that the tolerances for the
Elekta dMLC technique depended on the leaf speed to ensure that the dose error
did not exceed 2%. For example, if a leaf moves 50 mm in 20% of the total dose
prescribed, a position tolerance of 5 mm gives a maximum dose error of 2% and
so on.
Partridge et al (2000a) described all three systems and referred to the primary
papers in which they had been previously described. The integration time varies
between the three systems: for the Royal Marsden NHS Foundation Trust system,
it is 0.2 s; for the Amsterdam system 1.3 s and for the Manchester system 0.14 s.
The Royal Marsden solution was to record ten camera frames with an integration
time of 200 ms. Since each frame was locked to a particular accelerator pulse,
the precise number of delivered MUs corresponding to the measured frames was
precisely known. For the Amsterdam solution, in which a complete frame takes
about 1.3 s and each of the 256 lines are read out sequentially, the time is stamped
into the 257th pixel of a 257 256 array. The Manchester solution is to trigger
the image acquisition at known percentages of the delivered dose. Partridge et al
(2000a) demonstrated how important it is to quantify and eliminate all systematic
measurement errors and they described the technique for doing this in some detail.
They also described performance measures in terms of absolute errors in leaf
position per leaf per frame, means over this quantity and rms errors and total
rms deviations. The rest of the study summarized specific examples taken on the
three systems which demonstrated that the dMLC technique for all prescriptions
and for all fluence output rates and for all leaf speeds was working within the
manufacturers tolerance. Note that this is not a study of dosimetric verification.
Partridge et al (2000c) have evaluated the capabilities of another different
commercially available camera-based electronic portal-imaging system for IMRT
verification. The system used was the TheraView camera-based electronic portal
imaging detector (EPID) system (Cablon, NL). Because it was not possible to
interact with the hardware and software they attached a special set of lightemitting diodes which coded for the number of MUs delivered. The monitor
chamber circuitry of the Elekta linacs produces a digital pulse every 1/64th of
an MU and, in this way, each image was tagged with the cumulative dose by a
series of dots running down one edge. Software was written to decode these dot
patterns to yield the precise number of MUs corresponding to each image. It was
shown that the system could be used to monitor the dynamic leaf tracking during
a breast radiation therapy. By adding frames together, the integrated intensity was
also shown to be a good approximation of the incident fluence.
Partridge et al (2000c) also attached an anti-scatter grid to the scintillator
to ensure that calibrations made with a large field were also appropriate to
small fields and they showed that this was so by making measurements of the
radiological thickness of a target using both small and large fields. They also
122
showed that the stability of the imaging system was good over both short and long
periods of time. The overall geometrical accuracy of the system was determined
to be 2 mm with a dosimetric accuracy of 1.2 MU. It was shown that the major
geometric distortion in the system was a rigid-body rotation of 1.2 0.4 about
the optical axis of the system but that this could be corrected. The system was also
shown to develop afterglow but, given that this was present at both calibration and
measurement stages, the degradations cancelled out. In conclusion the system was
suitable for IMRT leaf-tracking verification.
Partridge et al (2001b) have performed megavoltage computed tomography
(MVCT) in cone-beam geometry using an amorphous silicon flat-panel portal
imaging detector prior to the delivery of IMRT. The same detector was then used
to measure transmitted primary fluence. This fluence can then be backprojected
through the CT model of the patient to form an effective input fluence map for
each beam which can be compared with the actual fluence map from the planning
system for verification.
Partridge et al (2002) extended this and have conducted a proof-of-principle
study for performing 3D in-vivo dosimetric verification of IMRT (figure 3.40).
An Alderson RANDO phantom was CT scanned, a horshoe-shaped target was
defined with an embedded OAR and a five-field, five-levels-per-field IMRT plan
was made at DKFZ using KONRAD.
A flat-panel imager was used at treatment time to make an MVCT scan.
Each projection dataset required correction for detector sag and an iterative
deconvolution of the significantly large radiation scatter. The same flat-panel
detector was used to create both individual images of the step-and-shoot segments
and summed exit images of beam fluence. After again correcting for scatter,
these images were used to reconstruct the plane of input primary fluence (see
also Partridge et al 2001b). Good agreement was found with the predicted input
fluence from the planning system.
The predicted fluence was then projected into the MVCT scan to calculate,
using Monte Carlo-generated dose kernels, the predicted delivered dose map.
When compared with the dose map from the planning system, the agreement was
3% in low-dose-gradient regions, 3 mm for high isodose lines. The small size
of the detector limited potential clinical implementation. Also the time taken to
gather the projection data from MVCT (17 min) and dose (from 120 MU) was
too great. The technique makes assumptions about the conversion to Hounsfield
Units, the exponential attenuation and the authors comment at length on the
potential systematic and random errors.
Woo et al (2003) have built an automatic verification system for monitoring
step-and-shoot IMRT field segments using portal imaging. The signal from a
BEAMVIEW portal-imaging system were extracted to a frame-grabber and PC.
Overlaid on this is the corresponding field shape from the CORVUS planning
system. Edge-detection software could then pick up errors. Warkentin et al (2003)
have presented a method for using a flat-panel portal-imaging detector to verify
the dose in IMRT. The essence of the technique is that the fluence measured by
123
Figure 3.40. A schematic overview of the adaptive radiotherapy process. If the patient
geometry measured by an on-line CT system is out of tolerance, a decision must be
made as to whether it is possible to simply re-position the patient using a rigid-body
transformation and continue with treatment or re-plan. Note that the on-line CT facility
permits a calculation of the delivered dose distribution. (From Partridge et al 2002.)
the flat-panel detector has to be converted into a measure of detected dose by

deconvolving two physical phenomena. The first is the dose kernel due to build
up in the material of the detector and the second is the glare kernel due to optical
scatter of photons. The first of these was created using Monte Carlo techniques
and the glare kernel was created using an empirical fitting of a double Gaussian to
experimental data. Also the flood-field correction has to be built in to the portalimaging measurement. They showed that this technique allowed a prediction of
dose that agreed to better than 2% with film measurements and the differences
between the measurements created by portal imaging and those predicted from a
treatment-planning system were much higher than this showing the errors in the
planning system.
Yin et al (2000) have made portal measurements of IMRT beams and
backprojected these through reconstructed CT images to create a map of
the relative delivered dose distribution. These were compared with TLD
measurements and verified the accuracy of IMRT delivery.
124
3.13.2 Other EPID designs

Mornex et al (2000) have developed IRIS, a portal-imaging system which is based
on large-area-pixel-matrix technology using amorphous silicon detectors. The
IRIS system is capable of producing up to ten images a second and therefore
monitoring the beam steering in almost real time for dynamic IMRT.
Keller et al (2000) have studied the properties of a SLIC EPID for
verification of IMRT delivered by the dMLC technique and have determined that
the applicability depends on the nonlinear response of the system. For a maximum
leaf speed of 2 cm s1 , they find that a sampling rate of 2 Hz is required, somewhat
faster than is presently available.
Ploeger et al (2002) and Sonke et al (2003) have developed a method for
geometrical verification of dynamic IMRT that makes use of a scanning EPID.
They show that motion distortion effects of the leaves are corrected based on
the treatment progress. The method was tested on three different dynamic
prescriptions. Specifically Sonke et al (2004) showed that, for the amorphous
silicon imager, the potential causes of motion distortion and motion blurring were
ghosting and the fact that the flat panel integrates over the frame time. Neither
effects were shown to be clinically significant and the position of a leaf moving
at 0.8 cm s1 was verifiable to an accuracy of 0.25 mm.
Chang et al (2001) have used a Siemens BEAMVIEW, camera-type, EPID
to rapidly make measurements for step-and-shoot IMRT. The recorded intensity
maps were then compared with computed maps visually in real time. In order to
do this, they had to develop a simple solution for image distortion, a method to
handle histogram differences and image-analysis tools.
3.13.3 Technical aspects of EPID imaging for IMRT
Evans et al (2001) have studied the effects of spatial and temporal sampling on
the recorded dose distribution of an EPID monitoring an IMRT delivery using the
dMLC technique. The following physical phenomena were modelled:
(i)
(ii)
(iii)
(iv)
the delta function pulsed nature of the linac radiation,

the time interval between radiation pulses,
the finite acquisition time for each image frame,
the time interval between image frames (thus defining the imaging duty
cycle),
(v) the spatial sampling size of the EPID,
(vi) the imaging point-spread function of the EPID.
They studied three IMBs under a variety of imaging conditions. These
were a spatially continuous hemispherical IMB, a spatially sampled IMB from
a CORVUS inverse-planning system and a spatially sampled IMB from the
technique to improve breast dosimetry using IMRT.
The magnitude of the error in recording distributions was a function of the
sampling properties. Evans et al (2001) showed that provided images could be
125
recorded in 0.33 s every 12 s interval the errors were below 2%. Detailed
explanations were provided to substantiate the nature of the errors. Sometimes
these were oscillatory with the periodicity relating to the sampling interval and
the phase relating to the offset between the start of irradiation and the start of
sampling.
3.13.4 Extraction of anatomical images from portal images generated

during IMRT
Fielding and Evans (2001) have described the use of a camera-based EPID to
make measurements of the moving leaves in a dMLC treatment. Image frames
were acquired at a typical rate of one every 2 s with an integration time of
1.5 s. These frames were then summed to produce an integrated fluence profile
to compare with the prescribed fluence profile. The portal images were calibrated
pixel-by-pixel using the previously published quadratic method and this enabled
maps of anatomical thickness to be produced which are useful for both dosimetry
and patient set-up.
Fielding et al (2002a) have shown that anatomical information can be
extracted from electronic portal images created using IMBs (figure 3.41). They
described two techniques. In the first, a series of frames was summed to produce
an integrated portal image of the delivered beam with each image corresponding
to the multileaves in different positions. Following this, a second set of images
was created in exactly the same way and summed but with a PMMA block similar
to the thickness of the patient in place instead of the patient. The first summed
image was then divided by the second image and this led to a visualization of
patient anatomy. A modification of this, involving a second calibration known as
the quadratic method and involving four calibration measurements each delivering
the IMB to a different thickness of PMMA blocks, led to somewhat improved
results. This technique was applied to imaging prostate and pelvic nodes using
the Elekta/Theraview system.
Fielding et al (2002b, c) have also applied the method that they developed
for imaging patient anatomy using a portal-imaging detector and an IMRT beam
to the Varian amorphous silicon flat-panel imager. The technique relies on adding
together images of the patient with the multileaves in different positions and then
calibrating for the variable input intensities. Calibration successfully removes the
intensity modulations of the beam so that the skeletal anatomy of the upper thorax
and neck region can be seen in the calibrated image. Fielding et al (2003a, b) have
shown that, with this technique, it has been possible to detect patient (phantom)
movement as small as 2 mm, leaf mispositioning of the order 2 mm and a 1%
error in MUs.
126
Figure 3.41. A clinical image of the pelvis extracted from an EPID measurement
created with modulated fields. The quadratic calibration method was used with manual
registration. (Reprinted from Fielding et al (2002c) with copyright permission from
Elsevier.)
3.13.5 Blocking-tray-level measurement

Xu (2001) has developed a device which goes in the blocking tray comprising a
film and a diode to make measurements of fluence distributions and a single-point
dose measurement for IMRT verification. For 107 IMRT fields, 95% of the point
doses were within 5% of the CADPLAN calculated doses and all fluence map
films visually agree with plots from CADPLAN.
Glendinning et al (2001) have developed a novel strip-ionization chamber
comprising ten opposing pairs of plates which are placed in the blocking tray of
the accelerator. Each subchamber is 80 mm 3 mm with 3 mm separation. This
chamber was used to monitor the positions of leaves during a dMLC treatment
when set-up to deliver dynamic wedge and longitudinal and lateral sinusoidal
intensity modulations. The same patterns were monitored using an EPID and it
was found that this demonstrated that the Elekta SLi dMLC system in its research
release had a systematic lag with respect to that prescribed of 0.67 s. Thomas and
Symonds-Tayler (2003) have developed a block-tray mounted ionization chamber
also for IMRT verification.
Eberle et al (2003) have built a test liquid ionization chamber with 20 20
pixels. This resides in the output port of the linac head and can measure MLC leaf
positions accurately and fast for static and dynamic therapy. The performance
of different liquids was evaluated, tetramethylpentane giving the highest charge
yield.
Ma et al (2003a) have used a film in the accelerator-blocking tray to measure
the fluence of a modulated and open field. Given that the detector response is the
same for both fields, it cancels out in the determination of the actual delivered
fluence through a Fourier division method.
127
Renner et al (2003) have described a new way to verify radiotherapy

including IMRT. Film was used to measure the entrance fluence from beams.
These films were so calibrated as to give the bixel-by-bixel MUs for a modulated
beam. Then a pencil-beam dose calculation was made using the planning CT
data to create an independent measurement of the dose in a patient that could
be compared with the planned distribution. It was recommended that the film
(with build-up) be in the blocking tray of the accelerator so measurements could
be made at any gantry angle in case the accelerator performance changed with
orientation. Alternatively, EPIDs could be used.
3.13.6 The two-level MLC
Greer and van Doorn (2000) have proposed a novel design for a dual-assembly
MLC. The collimator is, in all respects, the same as a conventional MLC except
that every other leaf pair is in one of two alternate banks which are arranged
one above the other. So, the upper bank, for example, comprises all the evennumbered pairs of leaves with spaces in-between and the lower bank comprises
all the odd-numbered pairs of leaves also with spaces in-between. The leaves
can be moved by motors in the same way as the conventional MLC. In the
configuration described, movement of the leaves in both the upper and lower bank
(see figure 3.42) creates the usual geometrical field shaping and, for example, if
all leaf pairs in both banks are closed, the field is completely shielded.
Conversely, if and when one bank is moved laterally perpendicular to the
direction of leaf movement by just one leaf width, the attenuating elements in the
two banks line up and the spaces line up. In this way, the collimator can be used to
create a slatted view of the geometrically collimated radiation field whatever the
position of the leaf pairs along their direction of travel. Now imagine the reverse
phasing in which, once again, the attenuating elements in the two banks line up
but they now take the positions that were previously taken by the spaces whilst
the aligning spaces take the position previously taken by the aligned leaves. This
enables the taking of a second slatted portal image which, when combined with
the first, yields a full image of the patient in the treatment position.
Greer and van Doorn (2000) observed that if the slit width is small, for
example 2 mm (and the corresponding leaf widths are also 2 mm wide), then
just the single-exposure slatted image shows most of the detail and the true portal
image can be recovered via an image-processing technique. Conversely, if the
leaves and spaces are wide, for example 10 mm, then a single exposure is not
good enough to show patient detail and two single exposures, one phased with
respect to the other as described earlier, must be combined to form the image.
Performance of this collimator was quantitatively evaluated by creating
experimental single slits of radiation with the attenuating elements having
different spacings, different attenuating thicknesses and being at different sourceto-collimator and source-to-detector positions. Greer and van Doorn (2000)
showed a large number of very detailed graphs showing how the profile of
128
Figure 3.42. (a) A schematic diagram illustrating the multileaf collimator design in the
shielding position. Each bank of MLC leaves (viewed end-on) is split into two vertically
displaced levels, with each level consisting of every second leaf. The leaves in the upper
level shield the leaf width spaces or slits in the lower level. (b) A schematic diagram
illustrating the multileaf collimator design in its imaging position. One of the levels is
moved laterally by a leaf width so that the leaves are aligned with the other level. Radiation
is then transmitted through the collimator as multiple-slit fields. (From Greer and Van
Doorn 2000.)
129
radiation delivered through a slit depends on these properties. They also

modelled these profiles by the well-known technique of determining the profile
of radiation as the sample of a dual Gaussian source seen by sets of points in the
detector space. In general, the analytically-modelled profiles closely matched the
experimentally-measured ones.
They ascribed the origins of their concept to the well-known patent by
Swerdloff et al (1994a,b) for a tomotherapy fan-beam intensity-modulating
collimator with alternate vanes in two vertically displaced banks. However,
the concept proposed is quite different being appropriate to the geometry of a
conventional MLC and the new proposal is not for tomotherapy.
A key observation was that, as the slit width decreased, the profiles become
broader than the geometric slit projection resulting in increased overlap of
adjacent profiles. They also studied a configuration in which the leaves in the
two banks were slightly larger than the spaces in the adjacent bank and could thus
remove the need for any machined tongue and groove.
They also proposed that lateral movement of one bank with respect to another
by a partial leaf width could be used to improve the spatial resolution of multileaf
collimation. As far as can be told, no device of this nature has been constructed.
The experiments reported in the paper were made with single slits although a
single bank collimator comprising poured lead slats had been constructed for
experimental purposes.
The periodicity inherent in dual exposure images even with a 10 mm slit
implies that adjacent slit profiles do not superpose according to the laws of
perfect superposition. This may cast some doubt on whether the Manchester grid
collimator will be able to collimate fields that are equivalent to those collimated
by a microMLC.
Chang et al (2003) have studied a four-bank MLC in two levels. The jaws
of the accelerator are each replaced by a microMLC and they show that this leads
to a better ability to segment modulated fields when compared with the use of the
conventional two-bank MLC.
3.13.7 Water-beam-imaging system (WBIS)
Li et al (2001a) have designed, built and implemented a verification system for
measuring delivered IMRT dose distributions using a water-beam-imaging system
(WBIS) (figure 3.43). This system comprises a 30-cm-diameter plexiglas cylinder
containing a 5-mm-wide single slice of plastic scintillator with 20 cm of clear
water on either side of the scintillating screen. The scintillator is backed using
reflective paint. The scintillator is viewed end-on through an optically transparent
window by a charged-coupled device (CCD) video camera. This camera can
capture 512 512 10-bit pixel images of the screen with an integration time
adjustable from 1 to 120 ms. The images are processed by a frame-grabber that
can sum between 1 and 2000 images. It was determined that the output response
of the WBIS was linear with delivered MUs. Background images are subtracted
130
Figure 3.43. A schematic diagram of the Wellhofer water beam imaging system (WBIS).
The data line from the control unit interfaces via a frame grabber board with a portable
personal computer. This system was developed at Stanford University School of Medicine.
Dose distributions recorded with it were compared with those created by Monte Carlo
planning techniques. (From Li et al 2001a.)
from each field measurement. The measured dead times were found to be less
than 1 ms at all delivery dose rates and therefore negligible.
The image captured by the CCD camera is blurred by the scattered light
represented by a point-spread function which was fitted to a multiple Gaussian.
Through a calibration procedure, the parameters of the multiple Gaussian were
determined. Then all images including those with background subtraction were
deconvolved using this multiple Gaussian function and the dose distributions so
captured by the device were then shown to agree much more closely with Monte
Carlo-predicted dose distributions than those without the deconvolution process.
The equipment was used to make measurements of the integrated dose from a
series of beam-intensity-modulated irradiations from a variety of gantry angles by
capturing the effect of each delivery and adding frames together to create a picture
of dose in a transverse-axial plane that was then correlated to the predicted dose
distributions from a planning system including a Monte Carlo dose-calculating
engine. Agreements between the deconvolved WBIS images and the prediction
of the Monte Carlo method were very good. The key feature of the device is
that data processing and comparison between processed measurement and the
reference prediction could all be performed within 510 min, much faster than
the use of films, BANG-gel or TLD measurements.
131
3.13.8 Integrated portal fluence and portal dosimetry

Chang et al (2000a) have used the Varian PortalVision Mark-1 System to record
frames of portal fluence as the leaves move during the dMLC irradiation. They
used a high (900)-MU measurement so that the leaves could be slowed down so
as not to move too far during the 9 s it takes to record and store a high-resolution
image. The portal image pixel values can be related to a measure of fluence at each
pixel. Then, by summing the images weighted by the time interval for each image,
a measure of the total integrated fluence at each pixel was obtained (figure 3.44).
This image comprised both primary and scattered fluence and Chang et al (2000a)
assumed a linear combination of these, weighted by two fitting constants. Then
they defined a space-varying error term as the mean-squared deviation between
the expected measurement and the actual measurement. For a so-called perfect
IMRT delivery, the fitting constants were determined to minimize this meansquare deviation.
This formed the basis for estimating the importance of genuine delivery
errors, for example those due to a stuck leaf, a wrong modulation or erroneous
jaw positions. They showed that the mean-square difference between expected
and measured integrated portal fluence was a good parameter to establish that an
error may have occurred and the action level for it. They also made absolute dose
predictions to act as a verification tool for the dMLC IMRT technique.
Chang (2001) has used the Varian aS500 amorphous silicon EPID for
verifying IMB delivery. Twenty-five prostate IMRT subfields were recorded
independently as images and converted to dose rates using a calibration curve.
These were then integrated over time to obtain a measured dose profile which
was then compared with the intended profile for treatment planning. Using a
15 MeV beam, the correlation between measured and intended dose profiles was
acceptable within 1% but strong energy dependence was observed for the 6 MeV
beam making the calibration curve deviate from a straight line for that energy.
Pasma et al (2001) have converted portal images into measures of transmitted
fluence and compared these with predicted fluence for the dMLC technique.
The technique developed has been implemented inside the Varian CADPLAN
treatment-planning system and the method was tested on a Varian 2300
CD treatment unit, showing that, outside narrow peaks and steep gradients,
measurements agreed within 2%.
Van Vliet-Vroegindeweij et al (2001) have used an EPID to measure beam
profiles and compared with those produced by a 3D treatment-planning system.
Kirby et al (2002) have made measurements of integrated exit fluence in
IMRT when delivered with the Elekta Precise linac and when measured with the
Elekta iView (fluoscopic) and iViewGT (amorphous silicon) portal imager. It was
found that summed fluence patterns agreed well with the fluence data extracted
from the inverse-treatment-planning system PLATO.
Van Esch et al (2001a, b) have shown that predicted portal dose maps agree
with the measurements from an EP10 (Varian PortalVision) to better than 3%.
132
Figure 3.44. A moving picture of an intensity-modulated beam for prostate treatment

using the sliding-window technique in use at Memorial Sloan Kettering Cancer Center.
Twenty-five images were taken during the course of treatment and eight of them are shown
here labelled 1, 4, 7, 11, 15, 19, 22, 25. Because the Varian Portal Vision is a dose-rate
imager, each image is converted to a dose-rate map, then multiplied by a weighting factor
corresponding to the time interval for that image. The weighted dose rate maps are then
summed to obtain the integrated dose map shown in the lower right-hand part of the figure.
(Reprinted from Chang et al (2000a) with copyright permission from Elsevier.)
Kim et al (2002) have predicted, using Monte Carlo techniques, the portal
image that would be created using an IMRT delivery for a clinical head-andneck case using dMLC delivery. They showed that the Monte Carlo simulations
successfully reproduced the measured pre-treatment and the transmission dose
images.
3.13.9 IMRT verification phantom measurements
A very large number of reports have been made of techniques to verify IMRT
delivery using phantoms.
Ma et al (2001a) have described techniques for verification of IMRT at
Stanford where IMRT has been implemented clinically since the end of 1997.
(a)
133
(b)
Figure 3.45. IMRT plan verification using a PMMA phantom with a bone insert (left) and
isodose distributions for the plan generated by CORVUS using 15 MV photon beams on a
Varian Clinac 2300 CD accelerator (right). (From Ma et al 2003a.)
Leaf sequences were verified using film or the WBIS. Monte Carlo methods
were also used to verify the dose distributions in the patient. Two phantoms
were constructed (figure 3.45), both cylinders of 30 cm diameter, one of PMMA,
the other of water. Monte Carlo and CORVUS dose calculations agreed with
measurements to within 2% for both but differed by 4% when bone and lung
heterogeneity inserts were placed in the phantoms (Ma et al 2000a, 2003a). The
water phantom must, of course, be irradiated in the upright position requiring a
complicated conversion of gantry and collimator angles to match the planning
of a horizontal cylinder. The advantage of the PMMA phantom is that it can be
irradiated horizontally on the couch.
Radford et al (2000, 2001) have designed an anthropomorphic IMRT
verification phantom. This phantom was designed to accept one insert which
would be in place when acquiring CT data for treatment planning and a second
insert, which would provide precisely placed TLD and radiochromic film for
measurement of dose from intensity-modulated fields. The phantom was designed
to be mailable and so is lightweight and waterfillable. It is part of the multiinstitution NCI cooperative clinical trial effort.
Oelfke et al (2001) have reported on the clinical IMRT technique at DKFZ
Heidelberg. Planning is performed with KONRAD. The plan is transferred to a
phantom and special detectors, buried in the phantom, are able to make point-dose
and volumetric-dose measurements to compare with predictions (figure 3.46).
Verification of 67 patient treatments showed an agreement between calculation
and experiment to 0.6% 2.2%. By November 2000 143 patients had been
treated and the number in 2004 greatly exceeds this. Rhein and Haring (2001)
134
Figure 3.46. This shows two steps of a dosimetric verification process for IMRT. The
left-hand image shows the IMRT phantom loaded with several films in different transverse
planes. On the right-hand side is shown one of the developed films whose optical density
is calibrated to absolute dose values. These are then compared to the predicted dose values
provided by the treatment-planning programme. (From Oelfke et al 2001.)
described how 170 patients had been treated with IMRT at DKFZ Heidelberg
using a Siemens PRIMUS linear accelerator by this later date.
Kermode et al (2001b) have verified HELAX-planned multisegment fields
using the LA48 linear detector array and compared measurements with predicted
linear dose distributions. In the majority of cases, the HELAX system modelled
segmented-field radiation treatment to an adequate degree of accuracy.
Agazaryan et al (2000) and Agazaryan and Solberg (2003) have developed
a dMLC technique and built a phantom for 3D dose verification. The
sequencer includes transmission, interleaf and intraleaf leakage and includes leaf
synchronization to reduce the TG effect. It was found that output-factor correction
for small fields was the most challenging phenomenon for dosimetry. Measured
dose at the position of each diode in the phantom was compared with calculation
of dose from the sequenced leaf positions.
Chuang et al (2002) have investigated the use of metal oxide semiconductor
field effect transistors (mosfets) for clinical IMRT dose verification. The mosfet
was stable over a period of two weeks to within 1.5%. The mosfet presented
a linear response to dose between 0.3 and 4.2 Gy. The percentage-depthdose
measurements measured with mosfets agreed to within 3% of those measured
with an ionization chamber. A phantom was constructed and IMRT plans were
delivered to the phantom comprising 81 possible mosfet-placement holes. It was
found that the measurements from the mosfet and from the CORVUS planning
system agreed within 5%, whereas the agreement between ionization chamber
measurements and the calculation was 3%. It was concluded that the mosfet
detectors were suitable for routine IMRT dose verification. Their main advantage
135
is their extremely small detector size (0.4 mm2 ) and they can immediately be reused. They can make multiple point-dose measurements and they can also be
used for in-vivo dose measurements. Some concern over the angular dependence
of early mosfets has been disposed of now that the design and characteristics of
new mosfets have significantly improved.
Cosgrove et al (2001) reported on the verification measurements made in
a phase-1 prostate and pelvic node IMRT trial at the Royal Marsden NHS
Foundation Trust (figures 3.473.53). Treatment planning was (at that time)
carried out with a NOMOS CORVUS Version 3 inverse-planning system and
the leaf sequences were interpreted using interpreter code developed by Convery
and Webb (1998). To verify a plan prior to the start of a treatment course, a
volumetric approach was adopted using the pelvic region of an Alderson Rando
phantom. The IMRT beams, calculated for the patient, were recalculated onto the
Rando phantom geometry and then delivered to the phantom which contained
sheets of XV film loaded within it. Once developed, the films were scanned
using a Vidar scanner and 2D isodose distributions were generated and compared
with the plans. The quality-assurance goal was to ensure that high isodose lines
agreed to within 3 mm and/or doses in low-gradient regions agreed to within 4%.
Additionally, 80 TLD chips were used to verify a treatment plan. These were
positioned in appropriate places within the phantom and agreement in both lowand high-dose regions was typically within 2% of that calculated. Additionally,
films strapped to the gantry of the linear accelerator showed that the fluence
distributions were reproducibly delivered on subsequent fractionation occasions
(figure 3.54). Further details of this trial have been presented by Adams et al
(2004).
Lee et al (2001b) have designed a home-made cylindrical phantom which
can contain a microionization chamber and a thimble chamber as well as
films placed transversely to make verification measurements for IMRT. A slab
accommodating TLD chips can also be incorporated. This system is very similar
to that described by Cosgrove et al (2001).
Saw et al (2001) replanned CORVUS-generated IMRT beams designed for
a MIMiC-based IMRT delivery on to two phantoms. One phantom was 7 cm in
diameter and 12 cm long and the other was a 12 cm30 cm30 cm phantom and
the third was a 22 cm 30 cm 30 cm phantom. It was found that the difference
between the replanned dose calculations and the measured dose distributions were
smaller when the phantom had a comparable size to that of the patient.
Short et al (2001) delivered IMRT computed with the Radionics X-plan
planning system with plans exported to a Siemens Primus linear accelerator. In
an autosequence 3050 subfields were delivered. These were checked using film.
Additionally, the fields were exported to be replanned on to a phantom containing
TLDs and measurements were compared with predictions. The agreement was
to better than 5%. Rose et al (2001) showed that, at the same centre, the use
of a water phantom could demonstrate good agreement between planning and
measurements.
136
Prostate and Pelvic Node Clinical IMRT Trial. A

typical transaxial section looks like this.
Figure 3.47. A typical transaxial section of the prostate and pelvic nodes illustrating how
the small bowel loops into the concavity creating a case for clinical IMRT. (Courtesy of
Drs Viv Cosgrove and David Convery.) (See website for colour version.)
Prostate and Pelvic Node Clinical IMRT Trial

330
30
90
270
180
Figure 3.48. The five gantry angles used for pelvic IMRT at the Royal Marsden NHS
Foundation Trust. (Courtesy of Drs Viv Cosgrove and David Convery.) (See website for
colour version.)
137
Prostate and Pelvic Node Clinical IMRT Trial.

Inverse planning generates isodoses like this.
Figure 3.49. IMRT planning spares the small bowel looped into the concavity shown in
figure 3.47. (Courtesy of Drs Viv Cosgrove and David Convery.) (See website for colour
version.)
Mazurier et al (2001b) have used the inverse treatment-planning system

HELAX Version 5.1 to generate intensity-modulated field segments for the
PRIMUS accelerator from Siemens. Film measurements showed that the leafposition accuracy was better than 1 mm. The monitor chamber linearity was
adjusted so that the difference between the dose-per-MU obtained for 1 MU and
for 200 MUs was just 3%. The difference between the calculated and measured
output factors was less than 3%. Full IMRT was verified by comparing calculated
and measured dose distributions in an axial plane of an IMRT phantom with an
agreement better than 5%.
Gluckman and Reinstein (2000) and Gluckman et al (2001b) have developed
a computer tool to correlate predicted and measured dose distributions and
provide systematic measurements of equivalence or lack of this. Gluckman et
al (2001a) have adapted an IMRT prostate phantom for full 3D dose verification
using polymer gel and also multiple-plane radiochromic film inserts.
Depuydt et al (2002) have extended the gamma index tool of Low (2001)
to compare calculated and measured IMRT dose distributions in a phantom.
Sorvari et al (2000) have commissioned a linear accelerator delivering
the dMLC technique. Calculations were compared with measurements in
anthropomorphic phantoms containing TLDs and it was found that results
compared well except in low-dose regions where the delivered dose was about
5% lower than that expected by calculations.
138
Dynamic MLC QA using a RANDO Phantom

Anthropomorphic, sliced phantom with inserts to load
TLD chips. Film can also be loaded between slices.
Figure 3.50. A section of the RANDO phantom arranged for TLD loading together with
x-ray images. (Courtesy of Drs Viv Cosgrove and David Convery.) (See website for colour
version.)
Cardarelli et al (2001) have evaluated the MED-TEC quality-assurance

phantom for IMRT.
Bakai et al (2002) have described QA for IMRT in which point
measurements were made with the PTW LA48 linear diode array or a very small
ion chamber. Letourneau et al (2004) have used the MapCheck 2D diode array
for IMRT verification and quantified its performance.
Leybovich et al (2002) have shown that the use of small ionization chambers
does not improve the accuracy of IMRT dose verification and that larger ion
chambers (0.6 cm3 ) are recommended for absolute dose verification.
Boehmer et al (2004) described the IMRT quality-assurance tests performed
at the Charite Hospital in Berlin.
Reports summarized here are, by necessity, something of a mixed bag
of experimental arrangements. Further reports are expected because IMRT
verification programmes have to be established in all centres beginning clinical
IMRT.
3.13.10 Verification by software techniques
Lee et al (2001c) have developed a virtual pre-treatment verification for IMRT.
Essentially a computer program models the leaf motion of the moving MLC and
factors in all the needed physical processes. The fluence thus generated is then
139
Dose verification using a RANDO Phantom
Patients IMRT fields can be recalculated on the RANDO

phantom CT data. The resulting dose distributions can then
be verified with TLDs and film.
(The two dose calculations will slightly differ due to variations in
patient/phantom sizes.)
Figure 3.51. Recalculation of intensity-modulated fields onto a RANDO phantom for

IMRT verification. (Courtesy of Drs Viv Cosgrove and David Convery.) (See website for
colour version.)
compared with that from the inverse-planning system. Agreement to within 2%

was obtained. This is surprising unless the physical processes are also included
into the planning system.
Moran et al (2001a) have developed a semi-automated verification system
for the dMLC technique. The log sequence files from a Varian 2100 EX system
with a 120-leaf MLC were used to create the expected and the actual positions
for each leaf. From these data, the leaf trajectories were calculated and the
expected and actual dose images were created. The images were then compared
with digitized films or with digital fluence images from a portal imager. Moran
et al (2001b) have compared the dosimetry of static conformal therapy, static
segmental MLC therapy and the dMLC technique.
Teslow (2002) has developed practical extraction and report language
(PERL) programmes within the CORVUS treatment-planning system to
summarize the outcomes of inverse treatment planning and to provide tools for
supporting confident patient treatment decisions.
Chen et al (2000b) have developed a verification process for MLC-based
IMRT which makes use of CT virtual simulation. Treatment planning was
performed using CORVUS and the planned fields were then translated onto a
virtual simulation workstation which is able to compute a digital reconstructed
radiograph in the beams-eye view of each field. A fixed MLC field with the
same boundary as this IMRT field was then also created for portal imaging so
140
TLD Dose Verification
Small cylindrical inserts

are used to position the
3 x 3 x 1 mm3 TLD chips.
(TLD 100, supplied by QUADOS, UK)
Figure 3.52. TLD dose verification for IMRT. (Courtesy of Drs Viv Cosgrove and David
Convery.) (See website for colour version.)
Film Dosimetry
Film Scan
Calculation
Figure 3.53. Film dosimetry for IMRT verification using a RANDO phantom. (Courtesy
of Drs Viv Cosgrove and David Convery.) (See website for colour version.)
141
Gantry Head Films
Figure 3.54. Film strapped to the gantry exposed to show the delivered fluence. (Courtesy
of Drs Viv Cosgrove and David Convery.)
that the field boundary can be verified for each field. A dose calculation was
performed using the intensity pattern and compared with that produced by the
inverse treatment-planning system. Agreement was to within 4%. The authors
claim that this makes redundant the need for in-phantom measurements to verify
IMRT.
3.13.11 Comparison of delivered modulated fluence profile with
plan-predicted modulated profile
The IMRT-planning process usually comprises three stages. The first and second
are the computation of fluence maps and the coupled calculation of dose using
these fluence maps. The third is the use of a leaf sequencer to give the pattern of
MLC leaf movements which will deliver the required fluence. The whole process
is far more complex than the delivery of traditional or even geometrically-shaped
CFRT and so its QA is a major issue. Most workers to date have concentrated
on making measurements of the delivered dose and compared these with the
calculated dose plans. Techniques have involved using an EPID, film, multiplepoint dosimeters and the WBIS system. Xing and Li (2000) pointed out that
such quality-assurance techniques are dependent on the combined performances
of both the leaf translator (or interpreter) and the functionality of the MLC
equipment. They have decided to view this quality-assurance problem differently
and to separate the two elements of the process. So, in their study, they did
142
not consider the QA of the leaf movements directly but, instead, concentrate on
whether the leaf interpreter will produce the required fluences to match those
produced by the planning system.
The interpreter translates a 2D map of bixel fluences into a set of leaf
positions. In doing so, it needs to take into account the specific equipment
limitations imposed by each manufacturers MLC. It also needs to take into
account the transmission through the leaves and head scatter. In essence, the
interpreter must generate a set of leaf patterns which will produce a 2D intensity
modulation in which the combined effects of both direct radiation and leakage
radiation yield the desired fluence pattern that was generated by the treatmentplanning computer. This is an iterative process. It relies on the use of a specified
dose model (see section 3.2.2). The crux of the work presented by Xing and Li
(2000) was to show that, when fluence modulations from the CORVUS treatmentplanning system were passed through an interpreter with a suitable dose model,
the computed fluence modulations differed by no more than 0.5% anywhere in
the field from the required prescription of fluences. They showed that this close
agreement was found for 20 specified patients and for prostate plans comprising
eight 2D modulations per plan. They then went on to show that there was also
an agreement within 2% between the fluence modulations so computed and the
measurements which were made with the WBIS system thus showing that the
agreement did not critically depend on the model for dose deposition.
These conclusions are the same as those from Convery et al (2000) and
Cosgrove et al (2000). It could be argued that in separating the QA into these
two components and only looking carefully at one of them, the process is less
rigorous than those based on the use of portal-imaging systems.
Xing (2001) further reported on the separation of the problems of quality
assuring the leaf movement and the QA of the dose distribution. In this work,
the leaf sequences were fed to software which simulated the motion of the MLC
and generated a new fluence map which was then compared quantitatively with
the reference map of the treatment-planning system. This approach was used to
validate the CORVUS and HELIOS treatment-planning systems and separated out
potential errors due to leaf motion from those due to other delivery effects.
Azcona et al (2002) compared IMRT plans made with a HELAX TMS
system with measurements for several test modulated fields and several clinical
cases. They concluded that fudge factors in the planning system need to reflect
the size of the field components. Errors can still arise due to inadequate modelling
of transmission, head scatter and output factors. Some of these modelling errors
are due to different assumptions made about the jaws between planning and
delivery. Planning assumed the jaws were set at the envelope of the summed
field components whereas Siemens Mevatron Primus IMRT delivery set the jaws
to the envelope of each field component in step-and-shoot mode.
143
3.13.12 Verification of canine and human IMRT using in-vivo dosimetry

Phillips et al (2000, 2001a) have emphasized the requirement for QA of all aspects
of the IMRT chain. Paraspinal and nasal tumours in dogs were treated using the
Elekta IMRT techniques. Biological endpoints were used in live dogs to assess the
dosimetry. The clinical models were purpose-bred dogs treated using an Elekta
SL20 accelerator. Of 15 dogs, nine had experimental paraspinal tumours and six
were controls. Twenty-one fractions were delivered with 4 Gy per fraction.
The validation of dynamic IMRT for head-and-neck tumours was as follows
(Parsai et al 2001a, b). CT contours were first defined using drawing tools in
PRISM. Planning was performed using projection onto convex sets (POCS) and
the macro-pencil-beam algorithm and delivery was, in the early years, by dynamic
delivery and later by step-and-shoot (Elekta changed their supported technique).
The data from POCS fed into a translator which was verified using a forward
dose calculation and provided the segments needed for linac delivery. IMRT
was delivered using two 6 MV Elekta SL linear accelerators fitted with MLCs.
It was found that, on average, there were about 100 segments per field, about
seven fields per plan, about 400500 MUs per beam and an average delivery
time of about 20 min. A linear diode array was used to check beam profiles.
Parsai et al (2001a, b) showed profiles of dose delivered to a phantom which
showed good agreement between measurements made with a linear array of
diodes and calculations of the dose profile. Measurements were also made with
films. Delivery of individual beams to phantoms determined the MU calibration.
Dog cadavers were also irradiated with TLDs in the spine and the readings from
TLDs and the calculations matched quite well. It was determined that there was
no need for Monte Carlo calculations. Dose measurements were made using
in-vivo dosimeters. The precision and degree of accuracy with which IMRT
can be delivered was assessed by creating highly localized dose distributions
to tumours near critical structures. In-vivo and in-vitro measurements were
made. Physical measurements of the dose distribution were made in live dogs
with targets adjacent to the spinal cord in the head-and-neck region and also in
deceased cadaver dogs.
Parsai et al (2001a, b) found that, as well as dose profiles measured
in cadaver dogs closely following the pattern of the predicted dose profiles,
measurements using TLDs contained in the critical normal tissue structures of
live canines also matched calculations within 3%. Overall the survival rate for
both control dogs and dogs receiving IMRT were found to be markedly different
favouring those irradiated with IMRT. After a years follow-up of IMRT versus
non-IMRT treated dogs, it was found that the control animals showed severe
normal-tissue damage, severe tetraparesis, compared to the IMRT group. This
veterinary work has led to a successful implementation of an IMRT programme.
This group also developed step-and-shoot compensators using a special tool
within PRISM. This tool optimizes beam weights for field segments and can
eliminate segments with small MUs. Overall it was concluded that they had
144
validated the dose-calculation model and also the means of delivery using the
dMLC technique.
Engstrom et al (2002) present what they believe to be the first TLD in-vivo
(human) dosimetry of IMRT. A naso-oesophageal tube filled with TLD rods was
inserted into the patient before treatment and data from eight patients and almost
299 measurements agreed well with the planned dose from the treatment-planning
system.
Higgins et al (2003) have investigated in-vivo diode dosimetry for the routine
QA of clinical IMRT. For the type of diode selected, it was found, for delivering
a wide variety of intensity-modulated fields to a water phantom, that diode and
ion chamber agreed within 2.5%. The variability between diode and expected
(i.e. planned) delivery on average was less than 5% with some outliers rising
higher. They then conducted in-vivo measurements using diodes and found again
that the ratio of diode to expected dose was generally within 5% but could not
be guaranteed to better than 10%. The rationale behind this development is
the intention to use standard patient diode systems employed for conventional
radiotherapy translated across to IMRT.
3.13.13 Polyacrylamide gel (PAG) dosimetry for IMRT verification
Polymer gels (or polyacrylamide gels [PAGs]) are gels which, when irradiated,
change several physical properties that are amenable to readout using medical
imaging techniques. They change their transverse magnetic resonance (T2 MR)
relaxation time, their x-ray linear attenuation coefficient and their ultrasound
acoustic properties. Lepage et al (2001) have proposed a method of using
Raman spectroscopy to measure the concentration of proton pools and so to
understand the processes which take place when monomer gels are polymerized
by irradiation.
3.13.13.1 Review
MacDougall et al (2002) have conducted a review of the precision and accuracy
of dose measurement in photon radiotherapy using polymer gel dosimetry. This
was not a simple literature review but instead aimed to put together the data that
would form a meta-analysis for certain questions concerning the accuracy of gel
dosimetry with respect to other techniques. The authors surprised themselves
by finding that there were very little data of a firm nature, most authors having
compared polymer gel dosimetry with film or TLD measurements or specific
planning calculations. The systematic review by MacDougall et al (2002) on
the precision and accuracy of dose measurements in photon radiotherapy using
polymer dosimetry was criticized by De Deene et al (2003). They made the
points that the authors of the report were not experts in the field and they felt that
the authors had made incorrect and misleading impressions about the accuracy of
gel dosimetry. MacDougall et al (2003) then replied that much of the criticism
145
was based on a misconception that exists around systematic reviews and evidencebased medicine. The review was not meant to be a topical review in the traditional
sense but one in which the accuracy of a particular technique was evaluated purely
in terms of the evidence produced by the peer-review papers.
Polyacrylamide gel (PAG) dosimetry has much promise and is being actively
explored by many groups. It is promising for IMRT verification. It is probably fair
to say it has not become a serious rival yet for other more traditional dosimetry
methods. Firstly we review new methods for reading out PAGs.
3.13.13.2 PAG readout techniques
Hills et al (2000a) have shown that PAGs change their x-ray attenuation and
so may be imaged with CT. This is simpler and cheaper than using magnetic
resonance imaging (MRI) readout. The dose response was linear and reproducible
and insensitive to temperature but has poor dose resolution. Hills et al (2000b)
applied x-ray CT readout to gels from stereotactic radiosurgery. The high-dose
region was CT scanned in 3 mm intervals with 3 mm slice thickness. To improve
image quality, 16 images were averaged per slice and a background subtraction
was performed. Then the dose images were imported into the treatment-planning
software and registered with planning CT images using image fusion. The two
distributions agreed to within 1.5 mm. A second application was made to proton
therapy to characterize depth-dose. Trapp (2003) has made an in-depth study of
gel dosimetry readout using x-ray CT techniques.
Mather et al (2002) investigated the changes in ultrasound acoustic
parameters of irradiated gels. It was found that using a fundamental frequency
of 4 MHz, a pulse repetition rate of 1 kHz and using the far field, the following
were observed. The inverse acoustic speed varied quasilinearly up to about 20 Gy
with a speed-dose sensitivity of 1.8 104 1 105 s m1 Gy1 before
reaching a plateau with further increase in dose (figure 3.55). The variation of
attenuation with absorbed dose was quasilinear up to about 15 Gy with acousticattenuation-dose sensitivity of 3.9 0.7 dB m1 Gy1 (figure 3.56). The
variation in transmitted signal intensity with absorbed dose was also quasilinear
up to about 15 Gy with inverse acoustic transmitted signal-dose sensitivity of
3.2 0.5 V1 Gy1 . There was considerable scatter on the curves and further
work was planned. However, ultrasound readout may provide an alternative to the
use of MRI, optical tomography or x-ray CT.
Oldham et al (1998) presented a new technique to improve the calibration
accuracy in BANG-gel dosimetry. They made use of a 16-echo CarrPurcell
MeiboomGill multispin echo sequence and used echoes 314 to determine the
sample value of T2 . By irradiating a large number of individual long vials
irradiated end-on to take advantage of the depth-dose distributions to obtain a
range of doses and increase the calibration accuracy, they came to the conclusion
that the error on dose was due to the error in the intercept of R2 (= 1/T2 )
at low doses and was dominated by the uncertainty in the gradient for larger
146
Figure 3.55. The inverse of acoustic speed of propagation in an irradiated gel as a function
of absorbed dose shown in a comparison with the MRI result of the dependence of the R2
relaxation parameter as a function of absorbed dose. Notice that the former is linear over a
much wider range. (From Mather et al 2002.)
doses. The following year, Baldock et al (1999) showed that using a two-echo
technique, the error was dominated by the error in R2 . Most recently Low et
al (2000a) have repeated the measurements made by Oldham et al (1998) but
using the two-echo sequence. In a detailed analysis of the errors in dose, it was
concluded that their results supported the conclusion of Baldock et al (1999) that
the limiting calibration precision is the voxel-to-voxel standard deviation. It was
concluded that systematic errors due to MRI scanning artefacts limited the use of
the technique to an overall error of roughly 0.2 Gy. Trapp et al (2004a) have also
studied error propagation in gel readout.
Oldham et al (2001) have shown how optical CT techniques can be used to
readout irradiated gels. It was found that high precision (better than 1.3% noise
at high dose) and a spatial resolution of 1 mm3 were achievable (figure 3.57).
Oldham et al (2003) have shown further developments to automate scanning
(figure 3.58) and Oldham and Kim (2004) have developed techniques to correct
for distortion (based on scanning a matrix of needles) and for scatter.
147
Figure 3.56. The acoustic attenuation of irradiated gel as a function of absorbed dose.
(From Mather et al 2002.)
3.13.13.3 Use of PAGs for IMRT verification

Groll et al (2000) have implemented IMRT using the Varian dMLC system and
delivered treatment plans to an Alderson phantom and BANG-gel phantom, the
plans being calculated by CADPLAN. Calculations and measurements were in
sufficient agreement that the technique proceeded to clinical implementation.
De Deene et al (2000) have investigated the chemical stability of monomer
and polymer gels and two different types of long-term instabilities have been
detected. Gel dosimetry has been validated as a 3D dose-verification method for
CFRT treatments in an anthropomorphic phantom. Future developments include
the development of lung-equivalent gels.
Gustavsson et al (2000) have used polymer-gel dosimetry to verify the
delivery of a C-shaped target dose distribution by IMRT. The treatment was
planned using the HELIOS software from Varian and delivered using the slidingwindow technique on a Varian 2300 CD accelerator with a 120-leaf dMLC. Good
agreement was found between the dose distribution recorded by the gel and the
corresponding data from the treatment-planning system.
De Wagter et al (2001) have used polymer-gel dosimetry to verify the
dosimetric accuracy of IMRT to typically 2% with a spatial resolution of 1 mm in
three dimensions.
Love et al (2003) have verified that IMRT of the breast delivers a smaller
target dose inhomogeneity than conventional tangential fields.
148

(a)
Rotation
F
B
Translation
(b)
C
A
D
(c)
Figure 3.57. Optical CT scanning apparatus for irradiated gels: (a) schematic view from
the top, (b) left-hand side of scanner, (c) right-hand side of scanner. Labels are: A =
gel flask, B = optically matched water bath, C = laser, D = reference photodiode, E =
lab-sphere-field photodiode assembly, F = baseplate. (From Oldham et al 2001.)
149
B
C
Figure 3.58. Modified optical CT scanner for irradiated gels incorporating automated
multislice acquisition. Components are: A = laser, B = mirrors, C = beamsplitter, D
= reference diode, E = linear stepping translation stage, F = optical bath containing
immersed gel dosimeter, G = vertical translation stage, mounting rotation stage and
suspending gel dosimeter in the optical bath, H = field photodiode, and I = X95 (structural
mounts) support. The two mirrors B on opposite sides of the water bath translate the laser
beam across the optical bath for acquisition of each projection. (From Oldham et al 2003.)
Novotny (2002) has used polymer-gel dosimetry for verifying dose

distributions delivered using a Leksell gamma knife.
Vergote et al (2004) used polymer-gel dosimetry to verify a clinical IMAT
treatment for whole pelvic radiotherapy. They obtained agreement to 3% in
regions outside high dose gradients.
Trapp et al (2004b) have made the first PAG verification measurements that
have influenced the choice of clinical delivery.
150
3.13.13.4 New PAGs

Gustavsson et al (2001a) have optimized the composition of a polymer
gel for 3D verification in radiotherapy. Acrylamide was exchanged for
2-hydroxyethylacrylate. Two series of gels were prepared in which the
concentration of monomer and the bisacrylamide (BIS) per total amount of
monomer was varied percentage by weight. It was found that the maximum
concentration of the monomers was determined by the solubility of BIS in water.
Increasing the fraction of BIS lowers the absorbed dose sensitivity but extends the
dynamic absorbed dose range.
Gustavsson et al (2001b) showed the comparison between calculated and
measured IMRT dose distributions where the latter was a volumetric measurement
with a new polymer gel using copper and ascorbic acid called MAGIC (Fong et
al 2001). Calculated and measured dose-volume histograms agreed. The MAGIC
gel can be used under normal levels of oxygen which is a major advantage since
excluding oxygen in the manufacture of other gels is problematic.
Gustavsson et al (2002) has evaluated the MAGIC polymer gel for IMRT
verification. IMRT was delivered using 6 MV photons from a Varian 2300 CD
linac equipped with a 120-leaf dMLC and a kidney-shaped target was defined
with OARs. It was found using the (Low) criterion that 95% of the points in the
volume of interest fulfilled the criterion at the 3 mm or 3% deviation level. The
gel was read out both at 2 and 14 days after irradiation using an MR sequence
with 32 multiple spin echoes.
Fricke gels have not entirely been abandoned. Gum et al (2001) have
developed a special phantom comprising Fricke gel to measure an IMRT
treatment plan generated by KONRAD with seven fields. The gel phantom was
then immediately read out using an MRI system and agreed with the plan to
better than 5%. Parker and Brodeur (2002) have used ferrous-sulphate-gel-based
dosimetry to verify IMRT. The gel was placed in a head-shaped phantom and
preliminary results indicated agreement with calculation to better than 10%.
3.13.14 Film as an IMRT dosimeter
Sohn et al (2000b, 2003) have studied the dosimetry of small fields, created using
the NOMOS MIMiC, of size 0.5 cm 0.5 cm, 1 cm 1 cm, 2 cm 2 cm
. . . 6 cm 6 cm. They measured the central-axis depthdose curves, the profiles
at depth and the output factors and compared these with the predictions of a
CORVUS planning system calibrated using a 4 cm 4 cm field. Measurement
were made with radiochromic film. It was found that the planning system
accurately predicted the dose per MU at depth for all field sizes greater than
1 cm 1 cm. However, disagreements were found for small depths (build-up
regions) and in the penumbra regions. The output factor reduced with field-size
reduction but was accurately predicted down to 1 cm1 cm but wrongly predicted
for 0.5 cm 0.5 cm field size. The implication for IMRT is not clear but it is
151
apparent that, when using small fields, the need to make measurements of dose
and to at least question if not disbelieve the predictions of planning systems is
advised.
Minken and Mijnheer (2001) have investigated the potential of radiographic
film for measuring output factors of small fields for IMRT. It was found that
radiographic film behaved more like a solid state detector for small fields than a
water-like detector and led to an over-response due to beam hardening. Mijnheer
(2001a) showed that the photon energy spectrum changes with depth and the film
is spectrum dependent. It is already known that film irradiated side-on has a
different response to film irradiated face-on and relative dose distributions are to
be preferred over absolute dose measurements with film. IMRT verification can
also be performed with TLDs, ionization chambers and in-vivo dosimetry. Portal
imaging is also being developed to verify IMRT with the latest development being
the use of flat-panel imagers. This group concluded that radiographic film in dose
verification of IMRT plans should be handled cautiously.
Martens et al (2002) have studied whether film is an acceptably accurate
dosimeter for IMRT. There are concerns that radiographic film has an energydependent response, that sensitivity increases with field size and depth due to
an increased contribution from low-energy photons and that the photon energy
spectrum also changes with position from the central ray. It was observed that the
response varied by no more than 3% as a function of field size with respect to a
standard 10 cm 10 cm field. The film response varied with the dose rate (and
so fractionation) due to the well known failure of reciprocity and an intermittency
effect. These variations can be as much as 9% of the OAR dose in an IMRT plan.
However, if a maximum of five beams contribute to a plan, then the overall effect
is good to 2%. It was concluded that film is an acceptable IMRT detector.
Ju et al (2002) have investigated the low-energy response of film as an IMRT
dosimeter. Film over-responds to low-energy photons that are more present in
penumbra regions and regions of low photon attenuation. They demonstrated
this phenomenon by making film measurements of pyramid and inverted-pyramid
fluence distributions showing differences between one-off film measurements
and the sum of film measurements for the constituent field components. They
made investigations with film both perpendicular and parallel to the direction of
incidence of the radiation. They recommended the use of thin lead filters that
improved the match between film measurements and ion chamber measurements.
Bucciolini et al (2004) implemented this suggestion and showed that with the lead
filters the dose-versus-density curve was independent of field size for both normal
and tangential irradiation. They showed good agreements between IMRT plans
and film dosimetry as a result.
Fayos et al (2001a) have delivered IMRT with the Varian dMLC system
with plans created using CADPLAN. Verification measurements were made by
irradiating films in a phantom using patient-specific IMBs. When the dose on a
particular plan was analysed, the low-dose gradient and high-dose gradient areas
were studied separately. Fayos et al (2001b) use Kodak XOMAT-V film and
152
ionization chambers to verify the movement of leaves on a Varian 2100 CD Clinac

accelerator for IMRT.
Nguyen et al (2002) have presented the characteristics of Kodak EDR2 film
which allows it to be used for IMRT verification due to its extended dose range and
Huang et al (2000) have also used Gafchromic film in the verification procedure
of IMRT.
Low et al (2002a) have described a system in which they use a relatively
inexpensive document and transparency scanner as a densitometer for routine
verification of IMRT. They compared the performance of this instrument,
which was never designed for this purpose, with that of a confocal scanning
laser densitometer and found that the maximum-mean and standard deviation
of the measured dose differences between the document scanner and the
confocal scanner were 1.48% and 1.06% respectively. They argued that this
system provides accurate and precise measurements up to an optical density of
two, sufficient for routine IMRT verification and thus widening the scope of
verification of IMRT to smaller community clinics.
Perrin et al (2003) have developed a software tool to compare the
measurements of film from phantoms irradiated with modulated beams with
the planning predictions. A variety of visual analysis methods were available
including isodose overlays, dose-difference maps, profile overlays, difference
isodoses and dose-difference histograms as well as the image. Schreibmann
et al (2004a) have also presented a software package for dose visualization in
IMRT.
3.14 Quality assurance (QA) of MLC delivery

In this section, attention is focused on the performance of the MLC itself
rather than on the verification of clinical IMRT delivered with the MLC (see
section 3.13).
3.14.1 Average leaf-pair opening (ALPO)
Zygmanski et al (2000) and Zygmanski and Kung (2001) have presented a method
to identify dMLC irradiation sequences that are highly sensitive to a systematic
MLC calibration error. The concept of an average leaf-pair opening (ALPO) is
defined. This is, as its name suggests, the mean value of the distance between
leading and trailing leaves weighted by the MUs for each segment where the
average is taken over all segments and all leaf pairs. The denominator calibrates
for the total number of MUs summed over all leaf pairs. The paper gives some
simple geometrical examples (see also Kung and Chen 2000a).
There are two sources of a systematic MLC gap error. The first is centreline
mechanical offset (CMO) and the second is the radiation field offset (RFO) in the
case of a rounded-end MLC (see section 3.2.3). It was shown that a systematic
gap error of x = 2RFO + CMO leads to a fractional average fluence error
153
of x/(ALPO + 2RFO + ) where is generally of the order of 1 mm. Using

this result, any set of 2D dMLC sequences, either step-and-shoot or dynamic, can
be used to compute, firstly the ALPO and secondly, for given RFO and CMO,
the corresponding fractional average fluence error. Close agreement was shown
between analytically calculated values of this error and numerical values of this
error. All that remains is to specify the threshold beyond which this error must not
grow in order to obtain a binary yes, no distinguisher between sequences which
are acceptable and sequences which must be rejected. This paper includes the
sentence that the term dMLC technique means both step-and-shoot and slidingwindow delivery technique. Zygmanski et al (2003a) capitalize on this concept
by designing dMLC patterns that are least sensitive to patient motion.
Zygmanski et al (2001b) suggested a tolerance of 2 mm for the Varian 52leaf dMLC treatments. There is a trade-off between having a leaf tolerance which
is too tight (which leads to very accurate dosimetry) and the consequent increase
in delivery time due to tripping the accelerator (see also Zygmanski et al 2003b).
3.14.2 Routine QA of MLC leaf movement
LoSasso et al (2001) have described in detail their comprehensive QA for the
delivery of IMRT with an MLC used in the dynamic mode in the Memorial
Sloan Kettering Cancer Institute, New York (see also LoSasso 2003). They
described specific QA procedures that are followed to assure the quality of
operations. These resulted from an incremental improvement to the overall
QA programme started when IMRT treatments commenced in 1995. It was
identified that the calibration of the gap between opposing leaves was a critical
component of QA. Very small gap errors could lead to large dose inaccuracies.
Two methods were used to identify encoders that exhibited excessive count
loss. Firstly, specific training modulations were delivered to film from which
it could be determined visually which leaves had motors that were going out of
control. A second technique used a software utility to exercise leaves through a
calibration cycle, then an exercise cycle and then a second calibration cycle and
the change in counts between the two calibration cycles assessed the ability of
individual leaves to maintain their calibration during dMLC operation. Diagrams
were provided showing how many motors had been replaced over a period of
time and geographically which motors of the specific sets had been replaced
(figure 3.59). It was noted that the main replacements occurred in the central sets
of motors, those that were used most regularly for IMRT purposes and therefore
over-exercised. Some 77 motors had been replaced over a five-year period to
September 2000. The changes in the calibration of the leaf coders gave a very
visual indication of a failing motor and the technique at Memorial Sloan Kettering
is to replace motors long before they cause specific dosimetric problems. This
study requires detailed reading to understand the specific QA checks.
In addition to these machine checks, specific dose-based verification was
made for specific patient IMRT treatments (for a general discussion of this topic,
154
Figure 3.59. The chronology of leaf motor replacement in three treatment rooms
at Memorial Sloan Kettering Hospital, New York. The full lines indicate the actual
cumulative motor replacements for each MLC and the broken lines show the shapes of
the curves that would arise from applying the current criteria for replacing motors prior to
1998. Time lines indicate the approximate dates of MLC installation, the first static MLC
treatment and the first dMLC treatment for each MLC. The motors that were most likely to
be replaced were those that were used in the dynamic mode for prostate treatments. These
were the motors close to the central axis because these fields rarely exceeded 10 cm in
length. This is a snap-shot picture as the situation was in November 2001. (From LoSasso
et al 2001.)
see section 3.13). One pre-treatment check measured point doses to 2D dose
distributions in a phantom for comparison with calculated results. The ratio of
measured and prescribed doses for approximately 400 patients was within 1%. It
was noted that individual points that were beyond this mean ratio were usually
lying in the TG regions or were due to small set-up uncertainties. In summary,
LoSasso et al (2001) pointed out the three-stage process of IMRT QA. The first
is routine acceptance testing and commissioning. The second is the use of a
routine QA process and the third is a patient-specific QA process to minimize
the possibility of human error.
Dirkx and Heijmen (2000b) have implemented the dMLC technique on the
MM50 Racetrack Microtron in Rotterdam. Inverse-planning techniques were
developed to create intensity modulations and interpretation techniques to develop
leaf trajectories that take into account scatter, penumbra, transmission and other
155
dosimetric effects. Before introducing the technique clinically, they went through
a careful process of QA following guidelines previously given by Chui et al
(1996) and LoSasso et al (1998a, b, 2001).
The way in which the control of the MM50 is operated was described in
detail. The MLC is calibrated by successively positioning the leaves at five
positions, 0, 7, 14 and +15 cm, storing the corresponding readings of the
potentiometers in a file. Then, to position a leaf at an intermediate position,
linear interpretation between the stored calibration values was used. It had
previously been demonstrated that the MLC control is precise in terms of staticfield definition to within 0.05 cm.
The first test to be carried out swept a 0.1 cm by 40 cm slit beam across the
field stopping the leaf pairs at 7 cm intervals, a technique known to generate a
bright line of radiation at right angles to the direction of the leaf travel for perfect
leaf travel (sometimes referred to as a garden fence test). It was concluded over
an 80-day period that the leaf positioning was stable to within 0.05 cm. A second
test swept a 0.4 cm by 10 cm slit across the beam and measured the dose deposited
at the centre of a uniform 10 cm by 10 cm square thus built up and it was found
that the variation in dose was only 0.4% pointing to the stability of the gap width
in time being better than 0.01 cm. When the same test was carried out with the
gantry at different orientations, the same observation was made indicating that the
movement of the leaves was not affected by gravity.
Specific intensity-modulated fields created for a patient with prostate cancer
were then measured using a portal-imaging device and again the variation in the
dose profiles over a long period of time was found to be less than 1% and, on
average, the leaves were positioned correctly within 0.01 cm. There was a small
increase in inaccuracy to 0.03 cm when the leaves were entirely travelling against
gravity at a gantry angle of 270.
Dirkx and Heijmen (2000b) then introduced deliberate interrupts to the
treatment and showed that these introduced changes in dose distribution, which
were less than 1%. They thus concluded that the acceleration and deceleration of
leaves could be effectively ignored. Overall the results of the QA investigation
showed that the dMLC technique executed with the MM50 Racetrack Microtron
was acceptable for clinical introduction.
Essers et al (2001) have commissioned the Varian CADPLAN (HELIOS)
IMRT planning system and linked it to delivery of the dMLC technique using
a Varian 2300 CD machine. Quality-assurance tests were performed, some
similar to those recommended by Chui et al (1996) and LoSasso et al (1998a, b,
2001). By sweeping small slits to define fields and deliberately accelerating and
decelerating leaves they determined that the accuracy of leaf positioning was
within 0.5 mm and the effects of acceleration and deceleration were negligible
with dose errors less than 2%. IMRT profiles were delivered on separate time
occasions with a stability better than 2% even when leaves travelled against
gravity.
156
Figure 3.60. The figure shows (a) an IMB delivered to film measured at 5 cm depth in a
6 MV beam and (b) a measurement using a diamond detector and calculated (line) dose
profile in the direction of the leaf motion at y = 55 mm. (Reprinted from Essers et al
(2001) with copyright permission from Elsevier.)
A number of different 1D and 2D profiles were delivered and dose

calculation compared with measurement (figure 3.60). The discrepancy between
the two depended on the values assumed for certain parameters. It was found
that the best agreement was obtained for a leaf transmission of 1.8% and a
leaf separation of 2 mm. Agreements were within 2% or 2 mm in regions of
high dose gradients. Conversely, the agreement worsened when manufacturer
recommended values were alternatively set.
157
Some deliberate attempts to create somewhat clinically unrealistic narrow

peaks led to larger 6% dose discrepancies. TG underdose was not specifically
excluded but tends to be minimized because all leaf travel times are constrained
to be the same.
Xia et al (2002a) have studied the precision of the correlation between the
MLC positions and the cumulative MUs for Varian-equipment-delivered IMRT.
It was shown that, when a profile was delivered with small MUs and/or at a
high dose rate, significant artefacts could be generated for test fields. This was
explained in terms of the sampling rate (20 Hz) and the communication time
lag between the control systems. Bayouth et al (2003b) have commented on the
relation between leaf-position accuracy and dosimetric accuracy and presented
film tests to determine accuracy.
Duan et al (2003b) determined that, for a Varian accelerator fitted with a
120-leaf MLC, the delay time due to communication with the dMLC was about
90 ms. This leads to discrepancies in dose delivered with IMRT, which increase
with dose rate and with the frequency of beam interruptions. Low dose rates,
slow leaf speeds and low frequencies of beam interruptions reduce the effect of
the delay and catch-up cycle.
Klein et al (2000) have presented techniques for assuring the quality of
IMRT delivered using multiple-segmented fields. The origin of this requirement
was their goal to deliver segmented MLC IMRT, so that the prostate could
be given a dose that was different from that to the prostate and seminal
vesicles combined. The comparison of calculated with experimentally delivered
irradiation showed the dose distributions differed by no more than 4%.
Ramsey et al (2001a) have made a study of the leaf-position error during
conformal-dynamic and intensity-modulated arc treatments. The characteristic of
their technique is that they make measurements of leaf-position accuracy taken
with leaf velocities arranging from 0.33 cm s1 . It was found that the average
position errors ranged from 0.030.21 cm with the largest deviations occurring at
the maximum achievable leaf velocity.
Low et al (2001a) have studied the performance of a Varian MLC for
delivering IMRT in both static and dynamic mode. In static mode, it was found
that the error introduced by field abutment mismatches was 16.7 0.7% mm1 at
6 MV and 12.8 0.7% mm1 at 18 MV. They also found that the leaf-positioning
accuracy was 0.08 mm and so concluded the static MSF IMRT technique was
accurate enough. In dMLC mode, they studied the degradation in smooth fluence
delivery at high leaf speeds.
Woo (2000) has used the Siemens BEAMVIEW portal imager to verify
IMRT segments delivered by the dMLC technique for QA purposes.
Vieira et al (2001) have developed specific modulation patterns which, when
measured using an electronic-portal imaging device, can indicate whether sliding
leaves have been mispositioned. Measurements were performed both with a
Varian 2300 CD Unit and a MM50 Racetrack Microtron.
158
Bruinvis and Damen (2001) have described the various QA tests that must
be made for a treatment-planning system for IMRT applications.
Olch (2001) has described the QA tests that have to be employed in a small
paediatric American clinic in implementing IMRT using the Varian accelerator
and the Nucletron PLATO treatment-planning system.
Chen et al (2002a) have developed a set of calibration tests for MLC leaf
positioning in IMRT. They find that when the gantry is at 90 or 270 with a
collimator angle of zero larger discrepancies than expected were observed due to
the effect of gravity.
Mavroidis et al (2000) have presented the results of the framework
BIOMED1 EC project Dynarad showing the variability of radiotherapy
techniques in use throughout Europe and illustrating the benefits and quality
requirements of the more conformal techniques such as IMRT.
Bate et al (2000) described how the post of Quality Control Officer was
established in the University of Ghent IMRT treatment programme. During the
start-up phase of routine IMRT treatments, it was considered necessary to have
such a rigorous quality-control programme to monitor errors in medical treatment
prescription, simulation, treatment planning, the treatment data transfer and daily
set-up.
De Brabandere et al (2002) have studied 12 patients whose IMRT treatments
were planned with CADPLAN/HELIOS. They then characterized the beams
in terms of their skewness and asymmetry. They attempted to see whether
these quantities were generally the same for all 12 treatments. The means and
standard deviations were found and then a QA tool was proposed whereby if a
plan deviated from these, it was suspected of error. They admit their proposal
is somewhat counterintuitive given that IMRT treatments are designed to be
customised to each patient.
Beavis et al (2002) have implemented clinical IMRT using the CMS FOCUS
treatment-planning system and delivering the treatments with a Varian 600
CD linac using a step-and-shoot method with the Millennium-120 MLC. They
describe the QA that is required and show that measured MU comparisons agree
within 1.5% with predictions when dosimetry unfriendly segments are removed.
Valinta et al (2001a) implemented IMRT using a Siemens PRIMUS
accelerator linked to a CORVUS inverse-planning system and they describe the
quality control procedures developed.
3.14.3 Modelling the effects of MLC error
Cho et al (2001b) have studied the influence of random field perturbations on
dynamic IMRT. They did this by assuming a Gaussian distribution of the position
of leaves relative to the expected true position and quantitated the outcome in
terms of changes to the dose distribution and dose-volume histogram.
Liu and Xing (2000) have used CORVUS to evaluate the effects of both
random and systematic errors of MLC leaf positions in IMRT delivery. They
Summary
159
studied three sitesprostate, nasopharynx and brainwith five patients for each
site. They found that the systematic errors were a larger influence on the dose
distribution than the small random errors which tended to smear out. They
determined that a leaf-positioning accuracy of 0.8 mm was necessary if the target
dose was to remain within 3% of its specified value.
Parsai et al (2003) have investigated the effects of random and systematic
beam-modulator errors in dynamic IMRT via a modelling study (figure 3.61). The
leaf positions of MLC leaves during dynamic delivery and also of jaws during
dynamic delivery were randomized using a Gaussian distribution of specified
width. During this process, the mechanical limitations of the particular MLC
(the Elekta MLC) were respected in the leaf-sequencing algorithm so that if
leaves were randomized to closer than 5 mm they were pushed apart by half
the required difference to establish half a centimetre gap. At the same time,
the maximum velocity allowed by the hardware of 2 cm s1 also provided a
constraint that was not violated during the randomization. Hence, the statistical
randomization led to an error distribution that was a clipped Gaussian rather than
a true Gaussian distribution. Beams were selected from a clival meningioma case
because of the close tolerances imposed by the sensitive normal structures. It was
shown that the target-dose uniformity decreased with increase in random errors.
When both MLCs and jaws were perturbed, the minimum target dose began to
deviate by 5% of the prescription when a standard deviation of 1 mm was used
in the randomization. In comparison, if MLCs or jaws alone were perturbed, 5%
deviation did not occur until the standard deviation was increased to 1.5 mm. In
addition to these random variations, systematic perturbations were modelled and
even errors of the order of half a millimetre were shown to result in significant
dosimetric deviations.
Somewhat counterintuitively, it might be expected that, if the results from a
large number of such randomizations were averaged, this statistical uncertainty
would all wash out but in fact this was shown not to be the case and target doses
were always less than those predicted without randomization. The authors offer
an explanation of this in terms of the fact that on average randomization of both
jaw and leaf positions will decrease the radiation aperture in 75% of the cases.
3.15 Summary
The MLC has an established place in the delivery of IMRT. Considerable
ingenuity has been applied to sequencers and to factoring in the effects of machine
constraints on delivery. Attention has been given to considering the measurement
of and role of leakage through the collimation and through the gaps in the
collimation. Both MSF and dMLC techniques are now widely established and
in clinical practice and we may expect these to seriously rival the use of the
NOMOS MIMiC, particularly in Europe. IMAT is now a commercially available
method and variations are appearing. Hybrid IMRT delivery methods have been
160
Figure 3.61. (a) Leaf trajectory for two opposing leaves in a dynamic delivery; (b)
corresponding fluence profiles. The dotted lines in (a) and (b) represent the state of the
MLC / diaphragms when perturbed randomly with a standard deviation of 1 mm. The bold
lines are desired deliveries. (From Parsai et al 2003.)
developed to include catering for the mix of MSF and dMLC techniques in one
delivery. Hybridicity also led to the use of rotating collimators to increase the
flexibility of MLC-based methods. There have been more microMLCs designed
and built as well as attention given to the theory of leaf width. A huge effort has
been given to QA of the fundamental equipment and also to techniques which
verify specific clinical deliveries.
Chapter 4
4.1 The Cyberknife

The Cyberknife is a robotic device holding an X-band linac, which is capable of
irradiating tumours stereotactically with feedback of organ motion to the robot
(see also chapter 6). The robot allows six-axis control. The use of the robotic arm
allows the linac to be moved with a high degree of freedom. Levin (2001) has
described in detail the scientific and business history of Accuray, the company
which markets the Cyberknife (figures 4.1 and 4.2).
The initial vision of the person, John Adler, who started the development
was that this would extend the concept of frame-based head-and-neck stereotactic
radiotherapy and radiosurgery to frameless whole-body radiotherapy and
radiosurgery. Adler had worked with Lars Leksell in Sweden, the developer of
the Gamma knife (now manufactured by Swedens Elekta AB Company). Adler
first worked at the Harvard Medical School in the Massachusetts General Hospital
in Boston. He then left Harvard in 1990 to go to the Stanford University School of
Medicine where he founded Sunnyvale, California-based Accuray with the goal
of carrying Lars Leksells vision from the brain to the rest of the body. The
development is described as the third major step in radiotherapy technology last
century (the first being the use of computed tomography data post 1972 to better
define radiation shapes), the second being the development of IMRT in the 1990s
and the third being the future of radiotherapy centred around the Cyberknife. The
main thrust of the approach is to take account of patient movement, breathing and
other involuntary movements so that the exact location of the tumour is at any
one moment tracked by the robotic linac. Adlers philosophy was that, because
radiosurgery works so well in the brain, there was no biological reason why it
should not have the same clinical application throughout the rest of the body. The
development of the Cyberknife became possible when one of Adlers patients
introduced him to a small California-based company that made portable linear
accelerators originally designed for industrial inspections, which were relatively
lightweight and therefore easily attachable to a robot. The financial history of
162
(a)
(b)
Figure 4.1. (a) The most recent version of the Cyberknife and its six motions (from
Accuray website); (b) also showing the x-ray image acquisition detectors. (Courtesy of
the Accuray Corporation.)
The Cyberknife
(a)
163
(b)
Figure 4.2. (a) shows the Accuray Cyberknife at Osaka, Japan. The 6 MV X-band linac is
mounted on the arm of a robotic manipulator and the treatment couch is positioned between
the two x-ray cameras and their respective diagnostic x-ray tubes. In (b) a schematic
diagram of the real-time tracking system is shown. By using the data from the infrared
tracking system and stereo x-ray imaging system, the robot arm moves the linac to the
ideal position in real time. This is an earlier version of the equipment shown in figure 4.1.
(From Shiomi et al 2001b.)
the Company is somewhat tortuous. Originally Adler looked for funding sources
through angel investors and by 1994 Accuray had five sales on the books and
had completed the installation of the first system. There was a period just after
this of (eventually abortive) negotiations with Picker International, a unit of
General Electric Co plc, and the recent success of Accuray has largely been due
to Asian-based financial investment mainly from Japan and Taiwan. Accurays
present Asian distributor is the Meditec Corporation, a wholly owned healthcare
subsidiary of the Marubeni Corporation.
The key feature, described as the crown jewels of the Cyberknife system,
is its use of minimally invasive on-the-fly imaging to continually feedback the
position of the tumour to the robot, now known as Accutrak or Synchrony
(figures 4.3 and 4.4). Essentially the procedure ensures that the tumour is in the
same place with respect to the robot at the beginning, during and at the end of the
treatment. Effectively the robot chases the tumour. The procedure is painless
but can last anywhere from 3090 min. Accuray received 510(k) approval for the
164
Cyberknife for treatments of head-and-neck tumours in July 1999 and in 2001

lodged an application with the Food and Drug Administration (FDA) for approval
to treat extracranial tumours such as those of the prostate, breast and liver. One
of the main difficulties of marketing this equipment is the difficulty of selling a
$3 million piece of capital equipment in todays cost-constrained environment,
particularly against larger competitors that are committed to the technology status
quo. (This is a general issue pertaining to the introduction of any new technology
which essentially challenges the role of existing technology. The argument is that
there can be no evidence for the benefit of the new until there are many clinical
installations and there are unlikely to be these until there is clinical evidence, a
Catch-22 situationsee also section 2.5). However, by the end of 2000, it was
reported that Accuray had sold and shipped a total of 18 Cyberknife systems,
five in the United States and 13 in Japan. This number is growing (Thompson
2003, private communication). Currently, European regulatory approval is still
to be sought. It is claimed that the main advantage of the Cyberknife is that
it can be used for both radiotherapy and radiosurgery with the equivalent of
intensity modulation whereas most photon IMRT systems can only be used for
radiotherapy. In particular, it is expected that with patients having wide access
to the internet, they will look for centres with this kind of equipment and the
marketing side of Accuray certainly believes that one cannot afford not to have a
Cyberknife.
Schweikard and Adler (2000) have described how patient motion can be
tracked and fed back to compensation using the robotic arm which controls the
Cyberknife system. The Cyberknife is equipped with two fluoroscopic cameras,
which constitutes an image-processing system and, based on the information
about the patients location, the robot chases the moving patient. The essence
of the technique is that stereo x-ray imaging determines the position of precise
internal markers which are correlated with the spatial location of external markers
placed on the patients skin which are viewed by infrared detecting systems
(figure 4.5). Then, during the treatment, the external markers are continually
monitored. The internal position of the target is computed from this monitored
position data and, if the target is not in the correct position with respect to the
beam, the patient or the beam may be realigned as treatment continues. Yan
(2003), using different equipment, have shown a similar reproducible correlation
between the motion of internal markers detected by x-rays and the motion of
external markers detected by infrared. Ramaseshan et al (2003) have also
correlated the movement of internal and external markers for lung.
Webb (1999, 2000a,e) has shown how robotic IMRT might become the
ultimate IMRT delivery technique in that it can sculpt contorted dose distributions
with high precision. He compared the technique with the delivery of IMRT
using more conventional methods and showed that, whilst these should remain
the clinical norm in the current climate of research into alternatives, it is useful to
know the degree by which these methods fall short of the ultimate.
The Cyberknife
165
Figure 4.3. A system overview of the Accuray Cyberknife and also the motion detecting
system. Infrared tracking is used to record the motion of external markers attached to
the patients abdominal and chest surfaces. Stereo x-ray imaging is used to record the
3D position of internal markers (gold fiducials) at fixed time intervals (e.g. 10 s) during
treatment. The relationship between the motion of the external markers and the internal
markers is determined and, from this, the continuous measurement of the motion of
external markers is used to translate to a pseudo continuous measurement of the motion
of internal markers. (From Schweikard et al 2000.)
Rodebaugh et al (2001a) have reported on the use of the Cyberknife for

treating lung lesions, some of which are partially mobile during the irradiation
time. This makes use of the Accutrak/Synchrony infrared light-emitting-diode
tracking system that is used to indirectly track target movement during respiration.
Rodebaugh et al (2001b) have examined the accuracy of the Accutrak system in
conjunction with the Cyberknife. It was found that the robot has an inherent
delay in responding to instructions from the measurement of position using skinlocated light-emitting diodes and the Accutrak uses an adaptive filtering algorithm
to predict the tumour location 0.8 s into the future. It was determined that the
means of the errors ranged from 0.6 to 3.4 mm. Murphy et al (2001) have shown
that, in feeding back motion to the Cyberknife, it is necessary to take account of
non-stationary motions and time-changing phase relationships.
166
Figure 4.4. A schematic diagram of the Cyberknife, illustrating (a) the X-band linac, (b)
the robotic manipulator, (c) the treatment couch, (d) the diagnostic imaging cameras, (e)
the x-ray sources. (Reprinted from Murphy et al (2003) with copyright permission from
Elsevier.)
Figure 4.5. This shows the way in which the Accutrak imaging system records the internal
movement of the tumour and feeds it back to the robot to correct beam direction. X-ray
sources A and B are viewed by x-ray detectors A and B (not shown, see figures 4.14.4).
These record the position of the fiducial markers every so often (e.g. every 10 s). The
infrared tracking system operates continuously recording the positions of the external
infrared markers. These positions of the internal and external markers are correlated and
used to predict the motion of the internal markers when only that of the external markers
is recorded. (From Schweikard et al 2000.)
The Cyberknife can be fitted with circular collimators of different radii.

Deng et al (2001b) have used the EGS4/Beam system for computing the elemental
dose distributions for various cone sizes. Ma et al (2001b) have reviewed
The Cyberknife
167
the role of the Cyberknife in radiation therapy emphasizing image guidance.

Specifically, detailed beam phase-space data are now available for accurate patient
dose calculation using the Monte Carlo method and treatment planning and beam
delivery have been validated using various phantoms and the dose distributions
have been verified using Monte Carlo simulations.
Guerrero et al (2001) have used the Monte Carlo package EGS4/MCDOSE
to make a prediction of dose distribution using the Cyberknife with and without
feedback of motion. When motion was not fed back to the robot, dose coverage
of thoracic tumours significantly decreased. With feedback of the motion, the
coverage was restored to close to the planned value.
Shiomi et al (2001a) have developed a method of QA for an image-guided
frameless radiosurgery system which uses radiochromic film. The QA method
was applied to the Cyberknife. In order for the Cyberknife to maintain high
accuracy, precise radiation is only possible if the coordinate systems, which the
Cyberknife keeps, i.e. the patients coordinates, the robotic coordinates and the
image-processing coordinates are all matched accurately. Before performing
irradiation in individual cases, Shiomi et al (2001a) calculate the overall
irradiation error using gafchromic film. The errors are then fed back into the
Cyberknife planning to improve accuracy. The technique comprises replanning
the patient treatment onto a phantom comprising a box phantom of a film set
between 3 mm-wide spacers. The treatment is then delivered to the phantom and
the films are read out to provide a measure of the dose distribution. The centre of
gravity of the dose distribution and that of the planned distribution is measured
for each slice and also in three dimensions and mislocation between the two is
assumed to be an error. A Cyberknife was installed at Osaka University Hospital
in April 1998 and has passed through three stages of commissioning. In its current
stage, the overall median error is known to be about 0.7 mm. Since the planning
is performed with voxels of size 3 mm3 and the range of brain movement within
the skull under normal conditions is estimated to be about 1 mm, the Cyberknife
was assumed to be adequately accurate. The particular Cyberknife at the Osaka
University Hospital was installed in such a way that the angle between the couch
and the robot was 45 . These measurements were made each time a patient was to
be treated. In August 2001 (Hemmi, private communication), there were just nine
Cyberknife systems in the world. The five in America are at Stanford, Cleveland
Clinic, the University of Texas, the University of Pittsburgh and Newport Beach,
CA and the four in Japan are at Osaka, Okayama, Yamaguchi and Kumamoti.
Shiomi et al (2001b) have also provided a description of the Cyberknife (in
Japanese).
Murphy (2002, 2004) has discussed three facets of frameless radiosurgery
using the Cyberknife. The first is the accuracy with which it can direct the dose
to the treatment volume. The second are the methods that can be used to deliver
radiosurgery doses to extracranial sites that cannot be easily immobilized by a
frame and the third is the pattern of movement that has been observed in patients
during frameless treatment.
168
Gibbs et al (2002) have presented the first clinical results of using the
Cyberknife system for treating spinal lesions. Gall and Chang (2003) have shown
that the dose calculation of the planning system coupled to the Cyberknife is
accurate to 2% provided tumours are not close to large tissue inhomogeneities. If
they are, the dosimetry can be compromised, an area for future investigation.
Related to the development of robotics and other special purpose
accelerators, a recent Medical Physics Point and Counterpoint argues the case
for and against a purpose-built linac for IMRT. The case for includes the view
that, for IMRT, large fields with flatness are not required since in general smallfield components make up the IMBs. Scanning low-energy linacs could suffice.
The argument against is based on the objection to a purpose-built machine, given
that conventional linacs with MLCs can do the job perfectly well.
4.2 The design of the shuttling MLC (SMLC)

Webb (2000c) proposed a radically new concept for an MLC for a photon linear
accelerator for delivering IMRT with high MU efficiency. The concept is to
consider each M (rows) N (columns) 2D IMB as a N/2 set of M (rows)
2 (columns) areas of modulation. Each area is then delivered by a set of M
shuttling attenuating elements (called here the shuttling MLC) with a very high
MU efficiency. The elements shuttle between each of the two rows comprising
the M 2 area and the modulation is provided by the dwell-time variation of
the elements (figures 4.6 and 4.7). The SMLC was extensively modelled and the
principles of this SMLC were discussed and examples illustrating its intended
operation were given. The main achievement reported was the development
and robust testing of an interpreter which describes the positiontime course of
movement of the elements as a function of MUs. This interpreter fully accounts
for leakage transmission through the elements. It completely avoids the acrossthe-rows TG underdose. A large number of 1D and 2D IMBs have been subjected
to this interpreter and it was shown that, for random patterns of fluence, the
SMLC is more MU efficient than the BortfeldBoyer technique (the most efficient
with a conventional MLC) when the modulation is highly structured. SMLCs
can increase the MU efficiency of IMRT compared with the multiple-static-field
(MSF-MLC) technique or dMLC technique with conventional MLCs.
In a companion study (Webb 2001a), two new arrangements similar to
that described in the earlier study, but with less mechanical complexity, were
shown to be constructionally simpler but less MU efficient (figures 4.8 and 4.9).
Additionally another new concept of SMLC was shown which also increases the
MU efficiency compared to the MSF-MLC technique and often improves the MU
efficiency compared with the previously reported SMLC for highly modulated
intensity distributions (figure 4.10). It also leads to zero TG underdose in the
direction orthogonal to that of the shuttling elements (so-called across-the-rows).
None of these designs have been turned into practical devices yet.
IMRT with the jaws-plus-mask technique
169
Figure 4.6. Showing the principle of the shuttling MLC (SMLC) or push-pull or
flip-flop intensity-modulating device. The device comprises multiple rows of attenuating
elements (such as U1) arranged in two banks, one immediately above the other. Just
one row is shown in the diagram and just four elements are shown for each of the two
banks. In practice, there would be many more elements per bank per row and many more
rows. The elements U1 and L1 can move into the spaces shown between them and U2,
L2 respectively. By varying the dwell time of the elements in these spaces and in their
positions as shown the intensity in bixels BIX1 and BIX2 can be modulated, e.g. U1 and
L1 can either be above each other as shown or both above each other in the adjacent space
or one can be as shown and the other in the adjacent space (essentially blocking both BIX1
and BIX2). The other pairs of elements behave independently the same way (e.g. U2 and
L2, U3 and L3 etc). Tongue-and-groove arrangements couple the rows (shown). Different
tongue-and-groove arrangements are needed at the faces of the elements along the rows
(not shown).
4.3 IMRT with the jaws-plus-mask technique

IMRT generally requires complex equipment for delivery. Just one study has
investigated the use of jaws-only IMRT with not discouraging conclusions.
However, the MU efficiency is still considered to be too low compared with the
use of an MLC. Webb (2002a) proposed a new IMRT delivery technique which
does not require the MLC and is only moderately more complex than the use of
jaws alone. In this method, a secondary collimator (mask) is employed together
170
Figure 4.7. This shows a 2D IMB and the technique to deliver it with the shuttling
MLC (SMLC). The 1D IMB has five pairs of non-zero bixels. Each pair is delivered
independently using an elemental shuttling arrangement of just two attenuators, one in the
upper bank and the other in the lower bank. These are shown shaded (and imply zero
fluence). Open elements with numerals indicate the open bixels and the MUs delivered
with the SMLC in this state. I p and I p+1 are the two adjacent bixels whose sum is the
maximum of the sum of all such pairs in the set. Here, e.g., bixels 7 and 8 require just two
states of the SMLC (component deliveries). The other bixels require three states. Note the
SMLC attenuators do not necessarily change at each monitor unit interval. They have one
of three states (the patterns shown) with a different number of monitor units assigned to
each state. The time taken to change states is ignored.
with the linac jaws (figure 4.11). This mask may translate parallel to the jaw axes.
Two types of mask have been investigated. One is a regular binary-attenuation
pattern and the other is a random binary-attenuation pattern (figure 4.12). Studies
showed that the MU efficiency of this jaws-plus-mask technique, with a random
171
Figure 4.8. 3D view of a HC(i)2-W-SMLC (HC = half complexity) with one static bank
and one movable bank plus an overall 1-bixel shift to move from irradiation phase 1 to
phase 2. The arrangement to deliver a two-dimensional IMB (6 6 elements are shown)
is shown.
binary mask, is more than double that of the jaws-only technique for typical 2D
IMBs of size 10 bixels 10 bixels and with a peak value of 10 MU (or quantised
into 10 fluence increments). For 2D IMBs of size 15 bixels 15 bixels with a
peak value of 10 MU (or quantised into 10 fluence increments), the MU efficiency
of the jaws-plus-mask technique with a random binary mask is almost triple
that of the jaws-only technique. Some further extensions to this concept were
presented showing that some more practical mask arrangements are possible but
with somewhat compromised MU efficiency. Some comments were provided on
practicalities and on delivery times.
Webb (2002b) showed that a much simpler relocatable mask than previously
conceptualized can lead to very similar improvements in MU efficiency and a
decrease in the number of field components compared with the use of jaws only
(JO). The new concept comprises a set of relocatable single-bixel attenuators
(SBAs) which can be moved into the field components otherwise collimated by
jaws only (figure 4.13). Typically for a 15 bixels 15 bixels 2D matrix of fluence
with a peak value of Ip = 10 MUs (or, equivalently, 10 stratified fluence levels)
and using just four SBAs the MU efficiency is nearly three times that of the JO
technique and the number of field components is reduced to about 0.6 the number
required by the JO technique. These gains become greater by using more SBAs
or for larger Ip values. Component reordering was achieved to minimize the total
delivery time including intersegment deadtimes. Practicalities were discussed.
For IMRT, a simple technique was described by Webb (2002c) which split
any N N matrix I of fluences into two mathematically equivalent matrices
IA and IB summing to 2I , each of which requires considerably fewer segments to
172
Figure 4.9. Principle of operation of a HC(i)2-W-SMLC (HC = half complexity) with one
static bank and one movable bank plus an overall 1-bixel shift. Shown is the arrangement
to deliver a one-dimensional IMB (six elements are shown corresponding to one row in
figure 4.8). The elements are labelled odd (O) and even (E). O M and E M are the odd and
even elements which have the greatest adjacent (odd + even) sum. The irradiation is in two
consecutive phases. For phase 1 (upper two parts of diagram), the static bank of attenuators
(part of a static bar when there is a 2D IMB) shields the odd bixels. The movable elements
shield either the odd or even elements in the configurations shown for the times ti shown.
The time t2 is the overall time for the first phase. For phase 2 (lower two parts of diagram),
the static bank of attenuators (part of a static bar when there is a 2D IMB) shields the even
bixels. The movable elements shield either the even or odd elements in the configurations
shown for the times ti shown. The time t7 is the overall time for the second phase. The
total irradiation time is T = t2 + t7 . All the times can be determined from the bixel values.
173
Figure 4.10. The figure shows the principle of operation of the 4-bixel-wide shuttling MLC
(4-W-SMLC). Just one component of the 4-W-SMLC is shown indicating the mechanism
which can generate four bixels of a one-dimensional IMB. This mechanism is replicated
in the direction of shuttling to create a row of such mechanisms which would deliver a
1D IMB. By stacking multiple such rows in columns, a 2D IMB can be delivered. In
the elemental mechanism, there are two elemental attenuators in the top bank and two
in the bottom bank. They are not shown to scale because they would be some 10 cm
deep and only 45 mm wide. The arrows indicate the possible directions of movement
of the attenuators. The elemental attenuators shown in position A can move discretely to
positions B or C. Those shown in position B can move to positions C or D. The movement
within the two banks is entirely independent. This generates 34 = 81 possible formations
but there is much redundancy (degeneracy). There are only six ways to create an aperture
with two open bixels, four ways to create a 1-wide aperture and 1 way to create a 0-wide
aperture (complete closure), i.e. 11 non-redundant formations. The degeneracy only affects
the total attenuation in the blocked bixels. In the 3D view, the attenuating elements are
shown shaded and the spaces into which they can move are adjacent. The gaps are for
clarity and not present in practice.
174
Conventional jaws
Mask pattern
Figure 4.11. Schematic diagram showing the jaws and the mask and the directions of
translation. The diagram is not to scale and, in practice, the mask would be a section of a
spherical annulus.
deliver IMRT using the jaws-plus-mask technique (Webb 2002a, b). When IA and
IB are delivered alternately at fractions throughout the treatment the total number
of field segments required is between one-half and two-thirds that of delivering
matrix I when unadjusted (see also section 6.3.19).
Earl et al (2003b) have provided a direct aperture optimization (DAO)
technique that can create JO IMRT with as few as 50 segments per field. This
number will depend on the matrix size, the number of fluence levels required and
the complexity of the modulation required.
4.4 The variable aperture collimator (VAC)

Webb et al (2003) extended these concepts of using a tertiary mask plus jaws for
delivering IMRT without an MLC. The new concept was to sweep a variable
aperture collimator (VAC) across the space of the IMB to be delivered and
to strip this IMB down into MSF components, each deliverable with the VAC
(figures 4.14 and 4.15). The stripping algorithm was described and it was shown,
for several designs of VAC, that the mean number of field components and mean
number of MUs was less using the VAC than would have been required for a
175
y-Jaw
aperture
The variable aperture collimator (VAC)
Secondary
collimator
(mask)
x-Jaw
aperture
Figure 4.12. A schematic diagram showing how field components are collimated by jaws
plus mask. The secondary collimator is shown as a matrix of black (attenuating) and
white (transmitting) squares. Each square corresponds to each delivered bixel. The four
locations of the jaws are shown as full black lines (made white in places for clarity). In the
arrangement shown, those bixels shown shaded are open in the secondary collimator and
thus irradiated. Note that this allows the grouping of an irregular shape of open bixels. The
exact shape depends on the translation of the secondary collimator.
jaws-only (JO) decomposition. The VAC would be simpler to construct than the
several previously suggested jaws-plus-mask (J+M) combinations. As well as
describing a simple VAC for the use with jaws, a design concept of a hybrid
VAC was proposed. It was also shown that adding the potential to rotate the
simple or hybrid VAC for some components relative to the field to be modulated
is advantageous. A DAO algorithm has also been worked out but currently only
for a 1D VAC (Webb 2004b).
176
5
6
4
8
7
5
5
8
1
8
7
1
9
6
0
8
0
7
0
5
3
5
6
5
7
5
0
2
0
5
6
6
7
9
0
6
4
6
6
9
6
5
3
1
7
3
5
5
6
5
6
5
Shielded by one Jaw
Shielded by 2 Jaws
Shielded by relocatable elements
Exposed to radiation
Figure 4.13. The concept of employing SBAs as well as jaws to create field components.
The jaws are shown collimating a 56 part of a larger 2D IMB (not shown in full). The key
indicates the functions of the jaws and the (in this case four) SBAs (relocatable elements).
In this example component, 1 MU could be stripped off to create the next residual matrix.
4.5 One-dimensional IMRT

Broggi et al (2002) have continued to use the Milan 1D IMRT system using a
beam modulator and have developed a rigorous and dedicated QA programme to
guarantee its accuracy and reproducibility.
Summary
177
Figure 4.14. Schematic diagram of one form of variable aperture collimator (VAC)
showing the six single-bixel attenuators (SBAs) in their parked positions (left) and one
possible component (right). Everything outside this is tungsten blocked. The VAC would
scan in a double-focused cradle (see figure 4.15). For each scan position, the six attenuators
could be moved by springs or push rods.
Figure 4.15. Prototype double-cradle arrangement for supporting the VAC. The figure
shows the cradle supporting a block of attenuator with three holes which would be replaced
by a VAC of the type shown in figure 4.14. The SBAs are double-focused. (See website
for colour version.)
4.6 Summary
The main pieces of apparatus for delivering IMRT are the NOMOS MIMiC and
MLC. However, the equipment reviewed here whilst playing a very small role,
178
or being only conceptual, represents an attempt at lateral thinking. It removes

thought from exploiting the MLC which was never intended for IMRT and thinks
ab initio. The MIMiC itself was of course the outcome of such lateral thought and,
apart from some concerns over matchline dosimetry, was and is very successful.
Time will tell the role of the other proposals reviewed here.
Chapter 5
Whenever a new medical technology is introduced there arises the question of

the availability of evidence for its need and efficacy (Bentzen 2004). Then
often follows the establishment of a Catch-22 in which the desired evidence
cannot rapidly accrue until there is widely diffused technology to gather the said
needed evidence. Added to this difficulty, the firm clinical evidence for improved
therapeutic ratio may come many years after the trials set-up to gather it. Worse
still, phase-3 clinical trials cannot even begin until phases-1 and -2 proof-ofprinciple trials have been completed. All these features conspire to produce a
patchy picture of evidenced-based medicine in the field of CFRT and IMRT. It is
this lack of large-scale hard clinical evidence that has fuelled some criticism of the
rapid implementation of IMRT and fired discussion (see chapter 1). Conversely,
it is this evidence that is exciting and fuelling further clinical IMRT.
The main clinical observations so far have been:
(i) There is lengthy observational evidence from Memorial Sloan Kettering
Cancer Institute for the improved normal-tissue sparing of dose-escalated
prostate CFRT and IMRT (Ling et al 1996) (see detail in section 5.1).
(ii) A positive outcome of the Royal Marsden NHS Foundation Trust phase-3
trial of prostate CFRT showed clear evidence that late rectal toxicity was
reduced with improved geometrical shaping (Dearnaley et al 1999).
(iii) Chao et al (2000) have studied 41 patients with head-and-neck cancers who
were enrolled in a prospective salivary-function study to attempt to correlate
the functional outcome of parotid-gland sparing with dose received by the
parotid during IMRT. It was found that the equivalent uniform dose (EUD)
concept suggested that the mean dose to the parotid gland is a reasonable
indicator of salivary-function outcome. The outcome of parotid function was
assessed objectively by measuring stimulated and unstimulated saliva flow
before and at six weeks, three months and six months after the completion
of radiation therapy. The study led to an observation that salivary output
decreased exponentially as a function of the mean radiation dose with an
exponential coefficient of approximately 4% per Gy. Hence, 50% of the
180
baseline saliva flow can be retained if both parotid glands receive a mean
dose less than 16 Gy.
Most studies aiming to demonstrate the improvements consequent on the
use of IMRT are based on demonstrating improved dose distributions.
Conversely, Chao et al (2001) have presented some of the first clear clinical
evidence for the reduction of late salivary toxicity without compromising
tumour control in patients with oropharyngeal carcinoma. Over a 30-year
period, up to the end of 1999, 430 patients with carcinoma of the oropharynx
were treated at the Mallinckrodt Institute of Radiology. Of these, 14 patients
were treated with the NOMOS MIMiC IMRT system post-operatively and
12 patients pre-operatively, definitively without surgery, using the same
equipment. Acute and late normal tissue side effects were scored according
to Radiation Therapy Oncology Group (RTOG) radiation morbidity criteria
with a median follow-up of 3.9 years. The crucial observation was that when
IMRT was compared with conventional techniques the dosimetric advantage
of IMRT did translate into a significant reduction of late salivary toxicity
in patients with oropharyngeal carcinoma. No adverse impact on tumour
control and disease-free survival was observed in patients treated with IMRT.
Indeed there was statistically insignificant improvement in two-year local
regional control.
It was found that, after IMRT treatment, only 1730% of the patients
had late grade-2 xerostomia. Chao et al (2001) gave a very powerful
presentation of why the blocking of salivary flow is extremely impairing
on the quality of life. Whilst the reduction or elimination of salivary
output is not life threatening, a severe loss of salivary flow is detrimental
to the processing of solid food leading to nutritional deficiency. It leads
to a deterioration of dental condition. Saliva facilitates speech and its loss
therefore hinders speaking and communication which may in turn cause the
patient to withdraw from social interaction. Also the deprivation of salivary
flow leads to opportunistic infections in the oral cavity. The prevention of
xerostomia is a major goal of head-and-neck IMRT and Chao et al s study
demonstrates hard evidence for the reduction in late salivary toxicity (see
also Chao and Blanco 2003).
(iv) Maes et al (2002) have also produced genuine clinical evidence for decreased
xerostomia following IMRT of bilateral elective neck irradiation. In 39
patients treated with conformal parotid-sparing radiotherapy, 78% have no,
minor or acceptable complaints of a dry mouth 612 months after treatment
and no recurrences were seen. Ipsilateral parotid gland sparing is more
difficult.
(v) Kwong et al (2003) have presented clear clinical evidence for reduced
xerostomia and increased alkalinity of saliva for late-response measurement
following patients who have received IMRT.
181
(vi) Patel et al (2002), Lee et al (2002a) and Mu nther et al (2003) have also
given clear clinical evidence of good response of salivary gland function
after IMRT for head-and-neck tumours (see also section 5.6.6).
(vii) Claus et al (2002) have irradiated 32 patients with sinonasal tumours using
IMRT and show that only one suffered dry-eye syndrome, a debilitating
consequence of conventional radiation that is, at best, an irritant and, at
worst, leads to total loss of vision. Mittal et al (2001) have shown that IMRT
significantly reduces toxicity in treating advanced head-and-neck cancers
and that the patient quality of life was significantly improved.
(viii) Yarnold et al (2002) have presented a preliminary analysis of a randomized
phase-3 trial between IMRT of the whole breast and conventional twotangent irradiations. It was found that a change in breast appearance was
scored in 52% of cases allocated standard 2D treatment and in only 36% of
patients allocated IMRT. These findings suggest that reduction in unwanted
dose inhomogeneity impacts on clinically observable late breast changes.
(ix) Mundt et al (2001) have presented their clinical experience with intensitymodulated whole-pelvic radiation therapy (see also section 5.5). Butterflyshaped concave high-dose volumes were created in the pelvis for the
treatment of gynaecological malignances and the absence of high-grade
toxicity in clinical follow-up was reported. Lujan et al (2001b) followed
up this study showing that IMRT reduces dose to bone marrow in
gynaecological radiotherapy. Mundt et al (2002) have designed either
five- or seven-field IMRT for whole pelvic treatment using CORVUS and
delivered plans using a step-and-shoot delivery on a Varian accelerator
(figure 5.1). Forty women have been treated and followed up for reporting
of acute toxicity. It was found that there was a significant reduction in
both gastrointestinal (GI) and genitourinary (GU) toxicity with no grade3 acute toxicity recorded at all. It was too soon to assess late toxicity
absence but this was expected to follow. One unexpected benefit of IMRT
was the reduction of irradiation of bone marrow leading to less severe
haematological toxicity. Consequent on this, IMRT provides a means to
increase the chemotherapeutic dose to these patients and so gain therapeutic
benefit in a concomitant way not involving the radiation.
Against this background and vastly outweighing the presently sparse clinical
data, there have been hundreds of papers published showing that CFRT and IMRT
generate improved dose distributions with more conformal tumour coverage and
sparing OARs. There are so many that no sensible list can be provided. However,
chapter 4 of Webb (2000d) summarized the clinical studies up to the year 2000
and there are over 1000 references in this work. Taken with the now wide
acceptance that biological models predict that better dose distributions will lead
to improved treatment outcome, these studies form the collective evidence for the
need for CFRT and IMRT. The studies range through tumour sites within breast,
182
Figure 5.1. Isodose curves from an intensity-modulated whole pelvis radiotherapy plan
superimposed on an axial CT slice through the upper pelvis. The small bowel and the PTV
are shaded in orange and green, respectively. Highlighted are the 100% (red), 90% (green),
70% (light blue) and 50% (dark blue) isodose curves. The figure shows a high degree of
conformity to the shape of the PTV in the upper pelvis minimizing the dose to the small
bowel. (Reprinted from Mundt et al 2002 with copyright permission from Elsevier.) (See
website for colour version.)
prostate, head-and-neck, brain, liver, lung, oesophagus, thyroid and paediatric

tumours.
Mohan (2002) has summarized the present state of some future directions of
IMRT research, noting that the key missing data concern the clinical outcome in
the long run which is subject to debate. Nutting (2003) points also to the future
for IMRT.
Schlegel (2003) has reported on the growth of clinical IMRT in Germany. An
audit at February 2003 showed that 651 patients had been treated at 11 centres in
Germany with DKFZ dominating this number at about 400: 60% of these patients
were head-and-neck patients, 26% were prostate patients and 14% made up the
rest. Schlegel has made an interesting observation that, if 20% of patients need
IMRT, this leads to a prediction of 40 000 patients per year in Germany whereas
the current capacity is 400 patients per year. He thus predicts that between 40 and
60 specialized centres will be needed. He points out the early developments in
IMRT were made in Stockholm, Heidelberg, Ghent and in London and yet now
Europe is behind the United States largely because of the fact that, in the United
States IMRT can be billed for at about three times the billing cost for conventional
IMRT of the prostate showing measurable clinical benefit
183
radiotherapy. In both Germany and the UK, there was no extra billing possible
for IMRT at the time of his review.
In the following sections, the latest clinical evidence is reviewed.
5.1 IMRT of the prostate showing measurable clinical benefit

One of the definitive implementations of prostate IMRT with the dMLC technique
is at the Memorial Sloan Kettering Cancer Center, New York. Between September
1992 and February 1998, 232 patients with histologically proven prostate cancer
were treated with CFRT. Of these, 61 were treated with geometrical shaping
alone and the other 171 were treated with IMRT. In both sets, the prescription
target dose was 81 Gy. The IMRT patients were planned with the DKFZ inverseplanning algorithm of Bortfeld et al (1994) modified by Mohan et al (1994). This
optimized a dose-based cost function with constraints and the possibility to have
an inhomogeneous dose distribution to the PTV. The IMRT technique was an
isocentric five-field method; in the overlap region of prostate and rectum, an 88%
prescription dose was set.
Zelefsky et al (2000) have published a report of the improvements
consequent on the use of IMRT for this large group of patients (figure 5.2). It
was found
(i) that the dose conformality improved with IMRT (not a surprising finding,
agreeing with others but parametrized carefully in the paper);
(ii) that there was a decrease in acute grade-2 and grade-3 GU toxicity but it was
not significant;
(iii) that there was a significant decrease in acute grade-1 and-2 rectal toxicity
with a concomitant increase in the number of patients having no toxicity;
(iv) that there was no difference in late GU toxicity but;
(v) that there was a highly significant decrease in late grade-2 rectal toxicity
(figure 5.3).
The study provided concrete evidence in support of the routine use of IMRT
for these patients. The evidence complemented and extended that found by
Dearnaley et al (1999) with respect to dose escalation in CFRT.
The study also escalated the dose to 86.4 Gy in an attempt to improve the
tumour control probability (TCP) further. The study of Zelefsky et al (2000)
also gave strong evidence to support the notion that rectal and bladder toxicities
exhibit a volume effect. It was also suggested that there may be advantage in
identifying specific subvolumes of the prostate which would benefit from a higher
dose, something that can be arranged with IMRT dose painting of dominant
intraprostatic lesions.
Zelefsky et al (2002) provided an update on the early toxicity and
biochemical outcome for high-dose IMRT of the prostate. Between April 1996
and January 2001, 772 patients with clinically localized prostate cancer were
184
Figure 5.2. Dose distributions of treatment plans designed for a prostate cancer patient.
The upper figure shows the outcome of conventional CFRT and the lower figure the IMRT
plot. Note the improvement in conformality of the PTV coverage by the 75 and 81 Gy
isodose lines in the IMRT plan. Also note that the 50 Gy isodose line avoids the femoral
heads in the IMRT plan. (Reprinted from Zelefsky et al 2000 with copyright permission
from Elsevier.) (See website for colour version.)
IMRT of the prostate showing measurable clinical benefit
185
Figure 5.3. Actuarial incidence of the late grade-2 and grade-3 rectal bleeding for patients
treated with 81 Gy IMRT compared with 80 Gy conventional 3D-CFRT. There were only
two cases of grade-3 rectal bleeding, one in each group. (Reprinted from Zelefsky et al
2000 with copyright permission from Elsevier.)
treated with IMRT at Memorial Sloan Kettering Cancer Center: 698 patients were
treated to 81 Gy, 74 patients were treated to 86.4 Gy. The median follow-up time
was 24 months. Of the patients 4.5% developed acute grade-2 rectal toxicity. No
patient experienced acute grade-3 or higher-grade rectal symptoms. Twenty eight
percent of the patients developed acute grade-2 urinary symptoms, one patient
experienced urinary retention and 1.5% of the patients developed late grade-2
rectal bleeding. 0.1% of the patients experienced grade-3 rectal toxicity. No
grade-4 rectal complications were observed. The data demonstrate the feasibility
of high-dose IMRT in a large number of patients. Zelefsky (2003) points to there
being no decrease in bladder toxicity.
In other centres, the evidence for the efficacy of IMRT of the prostate has
been gathering, as follows:
De Meerleer et al (2002a) have reported on acute rectal toxicity for 102
patients receiving IMRT for prostate cancer. They showed that acute toxicity
was most pronounced during the timespan of fractionated radiotherapy but that
grade-3 GI toxicity was not observed. Check-ups at 1 and 3 months after IMRT
rarely revealed grade-3 GU toxicity. Three months after treatment any toxicity
more than grade-1 was observed in less than 10% of the patients. This was not
a randomized controlled trial but is useful information justifying the continuation
of IMRT of the prostate.
186
Fiorino et al (2002) showed clear evidence of correlation between rectal

dose-volume histograms and late rectal bleeding for planning and clinical data
of 510 patients collated from four institutions, An important issue in validating
CFRT and IMRT concerns characterizing the dose distribution with dose-space
histograms and correlating these with either actual reported or probabilities of
complications. This is not easy for the rectum given it is a hollow organ. Fiorino
et al (2003) have modelled many situations with different geometries of the rectal
diameter, wall thickness and varying filling fractions. It was shown that the
normalized dose-volume histogram (normalized to equal numbers of calculation
points per CT slice) and the true dose-wall histogram correlated well for empty
or non-empty rectum. Given the relative ease to compute the former and the
difficulty of computing the latter, this is a very useful observation. Conversely,
for a very full rectum, the normalized dose-surface histogram (see Hoogeman et
al 2004) was more representative of the dose-wall histogram.
In multi-institutional trials of new ways to treat prostate cancer patients, CT
datasets will be contoured by many different clinicians. Foppiano et al (2003)
asked 18 observers to contour four patients and then constructed four-and sixfield conformal plans for each. The rectal dose was scored in terms of V40
through V75 to assess the impact of contour variation. This was judged less
than 4% after training and it was concluded that the benefits of increased patient
numbers from multi-institutional trials outweighed the disadvantages of relying
on different observers.
Hoogeman et al (2002) followed up 266 patients with prostate cancer who
were treated with either conventional or 3D CFRT. It was found that there was no
correlation between absolute dose-volume histograms and the incidence of rectal
bleeding but there was a volume effect for the relative DVHs of the rectal wall.
Kupelian et al (2001) have shown that late radiation damage due to IMRT
of the prostate is considerably less than that due to conventional CFRT when the
IMRT is delivered at 2.5 Gy per fraction to 70 Gy.
Teh et al (2001a) have treated 100 consecutive patients with localized
prostate cancer between February 1997 and January 1999 at the Baylor College of
Medicine, Houston, TX. They used the NOMOS MIMiC delivery system. They
followed up the patients in clinics at six weeks and every four months thereafter
and scored data on acute radiation damage using the RTOG acute-radiationmorbidity scoring criteria. It was found that, in general, acute GU toxicity and
acute lower GI toxicity were reduced with the use of IMRT. They also found that
the toxicities were all grade-2 or below and that the distribution between grades
0, 1 and 2 did not correlate with the mean dose above 65, 70 or above 75 Gy for
either bladder or rectum. They interpreted these results as indicating that modest
dose escalation can be performed for IMRT with safety. Teh et al (2001b) have
reported lower late GI and GU toxicity following IMRT for the prostate.
Raben et al (2002) have measured acute toxicity and quality of life for
patients treated with IMRT for prostate cancer. The overall rate of grade-1 and
grade-2 GU morbidity at completion was 41% and 45% and the overall rate of
Comparison of treatment techniques for the prostate
187
grade-1 and grade-2 GI morbidity at completion was 17% and 4%. The morbidity
decreased rapidly with time from the close of treatment.
Hofman et al (2002) have shown no major acute rectal complications when
delivering 76 Gy to the prostate with a three-field IMRT technique.
Livsey et al (2003) have shown that reduced / ratios for prostate suggests
that optimally selected hypofractionated regimes may produce better tumour
control and reduced late toxicity. They reported on the results for the first ten
patients studied in which no grade-3 toxicity was seen in a hypofractionated
regime. Mott et al (2004) described the development of class solutions for
hypofractionated prostate cancer treatment.
Clark et al (2003b) have conducted a randomized trial to investigate highdose hypofractionated multisegment CFRT to treat prostate carcinoma. IMRT
plans with three fields were created either using a 5 mm-leaf-width MLC or a
1 cm-leaf width MLC and prescriptions were either 74, 70 or 54 Gy to the prostate
GTV. Two segments were delivered at each gantry angle. The planning was done
by forward techniques.
Now we turn attention to some other clinical implementation of IMRT,
concentrating on specific organ irradiation but reviewing studies where clinical
outcomes were not specifically presented (with some exceptions due to difficulty
of systematically organizing material here).
5.2 Comparison of treatment techniques for the prostate

It is a fairly unsatisfactory process attempting to review worldwide experience in
implementing a clinical technique, particularly one such as IMRT of the prostate.
There are diverse planning systems, diverse IMRT delivery techniques, a plethora
of planning variables open to choice and a variety of institution conventions and
protocols. There is certainly no agreed protocol for treatment worldwide. As a
result any collation such as this is bound to be something of a mixed bag and it
is quite hard to draw firm conclusions. Certainly compared with the precision
with which one may review physics techniques, this is a more chequered field. As
much grouping as possible of similar investigations will follow.
De Meerleer et al (2000a) have taken a cohort of 32 consecutive patients
with different stages of prostate cancer and performed five planning studies on
each. All plans used just three field directions (figure 5.4). They compared:
(i)
(ii)
(iii)
(iv)
(v)
conformal geometrically-shaped radiotherapy (CFRT) optimized manually,

CFRT computer-optimized with no limit on PTV inhomogeneity,
CFRT optimized with a limit of 1520% on PTV inhomogeneity,
IMRT without a PTV homogeneity limit and
IMRT with inhomogeneity limit.
The results were averaged and average dose statistics reported. They also
reported PTV, rectal, bladder and femoral-head NTCP and the probability of
188
Figure 5.4. This illustrates an example of using segmented beams for IMRT as developed
at the University of Ghent. The aperture A spans the whole target volume. Aperture B
is the same minus a region in which the rectum is also in the line of sight of the anterior
beam. Aperture C is a third segment. The apertures are separately weighted (Reprinted
from de Meerleer et al 2000a with copyright permission from Elsevier.)
uncomplicated tumour control (P+ ) for the series. When calculating TCP, they
used some new data on the parameters and of the Webb and Nahum (1993)
model.
The overall conclusion was that the IMRT technique generated the best TCP
for the prostate, generated the lowest rectal NTCP and the highest P+ . Other
comparisons were presented in detail along with their significance or lack of it.
De Meerleer et al (2000b) further compared IMRT of the prostate using
three versus five beams . It was found that there were no differences concerning
the physical and biological endpoints for the target and that IMRT with three
beams better spared the rectum and bladder. The probability P+ of uncomplicated
local control was worse with five beams and, by carefully weighting the anterior
beam in the three-field technique, hot spots could be significantly reduced. They
concluded that only three modulated beams were needed and were at least as safe
as using five IMBs in the treatment of prostate cancer.
Sanchez-Nieto (2000) has shown how the use of TCP modelling can assist
treatment-plan optimization. She showed that IMRT with seven fields can reduce
the small-bowel irradiation when pelvic nodes are irradiated and that the NTCP
is only 11% compared to 34% with three-field conventional therapy and 29%
with three-field CFRT. This allows dose escalation to pelvic nodes. Because MRI
can show the actual tumour extent within the prostate, it is possible to increase
the dose to dominant intraprostatic lesions to 90 Gy delivering 70 Gy to the rest
of the prostate. Sanchez-Nieto (2000) analysed this using the TCP model and
189
showed that the TCP increase is entirely due to the dose escalation to the dominant
intraprostatic lesion and not due to general increase in dose to the rest of the
prostate.
Stasi et al (2000) have investigated the MSF technique for prostate cancer
using the HELAX treatment-planning system. A simple step-and-shoot technique
was performed using five gantry angles and just ten segments and this was
compared with a traditional plan created using the CADPLAN treatment-planning
system. The dose-distribution changes were remarkable. The dose delivered to
30% of the rectum in the IMRT technique was 10% less than that delivered using
a six-field conventional technique and a highly conformal PTV distribution was
obtained. The IMRT technique was benchmarked to measurements and found to
differ by no more than 1%.
Verellen et al (2000) have compared six different IMRT techniques including
(i) tomotherapy, (ii) IMRT based on a combination of the CORVUS planning
system and the Elekta MLC, (iii) static treatment beams with a miniMLC, (iv)
dynamic arc treatment with a mini MLC and (v) IMRT with a mini MLC.
Verellen (2001) has compared five different types of IMRT and their
appropriate planning systems. Verellen et al (2002) have compared four different
treatment strategies for the prostate: (i) sequential (NOMOS) tomotherapy, (ii)
dMLC or MSF IMRT with an MLC, (iii) dMLC with a miniMLC, (iv) dynamic
field shaping with a miniMLC (dynamic arc). The comparison not only looked
at conformality but also at treatment efficiency. The concern was that when all is
considered, IMRT may not be the preferred choice. The basis for the comparison
was a body of clinical experience in one clinic (AZ-VUB: Brussels). The specific
comparison focused on both a convex prostate and a concave prostate.
The NOMOS MIMiC was attached to a Siemens Mevatron KDS2 6 MV
accelerator. The dMLC and MSF technique was executed with the Elekta MLCi
attached to an SL15 accelerator. The miniMLC dMLC technique was executed
with the BrainLAB AG miniMLC attached to a NOVALIS accelerator and the
interpreter of Agazaryan et al (1999). The dynamic-arc technique also used the
BrainLAB miniMLC.
The MIMiC and macroMLC-IMRT techniques were planned with
CORVUS. The other two techniques using the microMLC were planned with the
BrainSCAN TPS, the dMLC using inverse planning and the dynamic arc using
forward planning.
For the convex target, all the treatment modalities met the treatment goals.
However, tomotherapy and the conventional MLC-IMRT technique needed a
tenfold increase in number of MUs and the miniMLC technique twofold when
compared with the dynamic arc method. The same trends were observed for the
concave target although then the dynamic arc technique did not meet the desired
dose reduction to the rectum.
Ezzell et al (2000) have developed a protocol for inverse treatment planning
for IMRT of the prostate. They use the CORVUS 3.0 planning system and tested
various prescription options including the beam arrangements. Interestingly, they
190
re-planned patients five times, each time with the same set of beam orientations
and tissue constraints but accessing beam directions in different random orders
to determine the variability in output of the CORVUS inverse-planning system.
The following results were found. A set of dose-volume-histogram prescription
values for the target and normal structures could be found that reliably satisfied
the minimum dose requirements. On average, it was found that a five-field plan
with laterals, posterioranterior (PA) and shallow anterior obliques provided 5%
more rectal sparing than other IMRT techniques that included a large number of
variations of using six fields or five, seven, nine or fifteen fields in equal angular
increments. They also found that, by reducing the allowed dose to the rectum
from 70 to 40 Gy, improved rectal sparing could be obtained at the cost of less
target-dose uniformity and that the best prescription was 60 Gy to the rectum.
Interestingly, it was found that for a given patient and beam arrangement, different
computer runs produced quite different solutions. For example, the dose to 30%
of the rectum varied by 9% for different computer runs of the optimized five-field
plan for one particular patient. However, unequivocally, the IMRT plans were an
improvement on conventional treatment planning.
Valinta et al (2000) have also used the CORVUS inverse treatment-planning
system to optimize a combination of static MLC-shaped beams for IMRT. They
used a Siemens PRIMUS accelerator and claim that angle selection is more
important than the number of intensity levels in sparing the rectum. They used
a limited number of fields and a limited number of levels of intensity. Valinta et
al (2001b) measured the leaf calibration accuracy to 2 mm and found that doses
measured and doses calculated by the treatment-planning system agreed to 2% at
the treatment isocentre.
Shentall et al (2001) have used the Nucletron PLATO treatment-planning
system to create five-field conformal and five-field IMRT treatments. The IMRT
plans were produced with two aims. The first was to achieve a PTV dose
uniformity of 5% and the second aimed at sparing rectum as much as possible.
The IMRT sequencing was limited to a total of 40 segments. It was found that
there was very little difference between the conformal and IMRT plans where
the main IMRT aim was PTV homogeneity. However, when the IMRT aim was
to spare the rectum, this was achieved with some significant gains. The overall
comment was that there is a considerable variation in the clinical results and such
comparisons may require to be patient specific.
Mayles et al (2002) have also implemented IMRT using the Nucletron
PLATO inverse treatment-planning software and delivered with an Elekta linac
using the step-and-shoot technique. IMRT was implemented for a prostate boost
to 70 Gy to the prostate GTV and 64 Gy to the prostate and seminal vesicles in
32 fractions. QA tests showed that changes were needed to correctly model the
collimator and phantom scatter because PLATO uses the collimator scatter value
of the largest field rather than allowing for the individual beamlets. When these
corrections were made, it was found that measurements in water phantoms and
in solid-water phantoms agreed with calculations to generally within 3% when
191
the effects of adding together separate modulated fields were included. Individual
modulated fields could differ by more than this 3% from the calculation. The
first patient was treated in February 2002 since when the Clatterbridge Centre
for Oncology has become one of the sites to implement prostate IMRT using
completely commercial solutions (Wavelength 2003b).
Young et al (2000) have developed a seven-field IMRT technique for
radiotherapy of the prostate which is specifically designed to spare the bulb of
the penis. The technique involves the use of variable margins to the prostate and
was compared with a 3D conformal technique employing seven treatment portals.
It was found that the mean dose could be reduced from about 40 Gy to about
25 Gy if a 1 cm target margin was used and that this did not compromise the
target homogeneity.
Sethi et al (2002, 2003) produced CFRT and IMRT plans for ten patients,
the latter plans made using tomographic IMRT and step-and-shoot IMRT. The
mean doses to the proximal penile tissue reduced by about 40% depending on
prescription dose level to the prostate which was escalated from 73.8 Gy to
90 Gy. The mean doses to the corporal cavernosa and corpus spongiosum (bulb)
similarly decreased by above 40%. Buyyounouski et al (2004) have also shown
the ability of IMRT to reduce dose to the penile bulb and erectile tissues without
compromising the target volume. This should help preserve sexual function
following high-dose radiotherapy.
Xia et al (2001) have studied a case requiring dose escalation to two
dominant intraprostatic lesions to 90 Gy whilst simultaneously delivering 75.6 Gy
to the remaining prostate. Three techniques were compared. The best dose
conformality was achieved by a sequential tomotherapy (MIMiC) plan created
using CORVUS. The inverse-planned segmental MLC (SMLC) plan gave the
lowest dose to the rectal wall. A forward plan using the UMPLAN system gave
the best dose homogeneity to the PTV. It was concluded that all three techniques
kept the rectal-and-bladder doses to below the RTOG grade-2 complication rate
of 10%.
Kung et al (2000a) have studied the use of IMRT to improve rectal sparing in
treatment of the prostate for patients who have a metal prosthetic hip replacement.
Plans were prepared using nine equally spaced 6 MV coplanar fields each
avoiding the prosthesis using the CORVUS inverse-planning system. These
showed superior dose sparing compared with geometrically-shaped conformal
plans made with the PLUNC treatment-planning system.
Luxton et al (2000) have studied the dosimetric advantages of IMRT for
treatment of prostate cancer for extended fields that include pelvic lymph nodes.
The dose to the lymph nodes, prostate and seminal vesicles was 50 Gy and this
was followed by a prescribed IMRT boost to bring the minimum dose to the
prostate to 70 Gy. The CORVUS inverse-planning system was used to create
the plans for IMRT and the Computerised Medical Systems (CMS) FOCUS
treatment-planning system was used to generate a 3D conformal treatment plan
using the same data sets. Twenty patients were studied. It was found that
192
IMRT can be used to provide significant bowel as well as bladder and rectum
protection while treating the prostate seminal vesicles and pelvic lymph nodes as
compared to conventional CFRT. However, this was achieved at the cost of less
dose homogeneity within the target. Luxton et al (2004) showed that reduced dose
to OARs led to reduced NTCP in prostate IMRT compared with 3D conformal
therapy.
Among shorter reports are the following:
Van der Heide et al (2001) have described their micro boost approach to
IMRT of the prostate.
Harnisch et al (2000) have shown that short-course IMRT can irradiate
the prostate gland to 65 Gy whilst simultaneously irradiating dominant
intraprostatic lesions to 80 Gy in 28 fractions. This can be done with rectaland-bladder doses limited to levels consistent with currently used schedules.
ODaniel (2001) has added further planning evidence that IMRT of the
prostate leads to clinical advantage compared with fixed-field CFRT.
Aletti et al (2001) have commenced IMRT of the prostate in Nancy, France.
Mihailidis and Gibbons (2002) have compared five- and nine-field IMRT and
shown that the choice between the two is patient dependent and depends on
femur and small bowel dose-volume histogram analysis.
Liu et al (2002) are using tissue typing of ultrasound prostate images to stage
the cancer for IMRT.
Zietman (2002) makes compelling arguments for the phase-3 randomized
trial of dose escalation of prostate cancer.
5.3 Royal Marsden NHS Foundation Trust pelvic and other

IMRT
Dearnaley (2000) has reviewed progress towards the use of intensity-modulated
techniques for reducing treatment-related side effects and other morbidities in
treating prostate cancer. Specific interests include boosting the dose to dominant
intraprostatic lesions and also treating pelvic lymph nodes. Nutting and Dearnaley
(2001) have given an update on the past, present and future Royal Marsden NHS
Foundation Trust trials on prostate CFRT and IMRT.
The Royal Marsden NHS Foundation Trust and Institute of Cancer Research
have implemented IMRT for treating prostate cancer with involved lymph nodes
(Adams et al 2004). Dose escalation to the prostate whilst simultaneously treating
the pelvic nodes with IMRT is also justified by Cavey et al (2004). The technique
used at the Marsden (Sutton) was the Elekta dMLC method in dynamic (research)
mode in the first instance and changed to a Elekta step-and-shoot mode after
the first ten patients. This stage of the disease is hard to treat radically due
to the presence of the dose-limiting small and large bowel which is enclosed
in the horseshoe-shaped PTV. CT scans were acquired at 5 mm intervals and
Royal Marsden NHS Foundation Trust pelvic and other IMRT
193
Figure 5.5. Typical beam intensity modulations derived from (a) CORVUS and (b)
HELAX TMS for IMRT of the prostate and pelvic nodes. It can be seen that the
CORVUS-generated IMB is very noisy and may not be physically realizable without
a great deal of careful interpretation using both leaves and jaws. Conversely the
HELAX-generated IMB is designed with deliverability checked during the iterative
optimization. (Reprinted from Adams et al 2004 with copyright permission from Elsevier.)
contoured to specify the bladder, rectum, femoral heads and small- and largebowel as OARs. A treatment technique was established using five fields planned
on CORVUS for the dynamic technique and on HELAX for the step-and-shoot
technique (figure 5.5). These were PA, two laterals and two anterior obliques
at 30 to the anteriorposterior (AP) line. Doses of 70 Gy to the prostate and
50 Gy to the pelvic nodes were prescribed with the dose to the nodes later
escalated to 55 Gy, then 60 Gy as part of a phase-1 study. Planning was tricky
but acceptable distributions were obtained (Webb et al 2001). The modulations
generated by CORVUS were passed into the interpreter developed by Convery
and Webb (1998) to derive the delivery sequences (Convery 2001). Those from
HELAX generated MLC leaf patterns directly. Before the treatment started, the
patient-specific fields were replanned onto an anthropomorphic phantom (see
section 3.13.9) at relevant locations and the doses recorded with 80 LiF TLD chips
and film between the slices. The TLD doses agreed with calculations to within
2% and the films were digitally scanned and also checked against the planning
distributions. Films were also attached to the exit window of the linac head to
assess the reliability of the leaf movement (Cosgrove 2001). The first patient was
treated in September 2000.
Adams et al (2002) have shown that the QA for IMRT can be reduced
following observation of successful comparison between predictions and
measurements for the first few patients in a phase-1 clinical trial of IMRT of
prostate cancer with involved pelvic nodes. Adams et al (2003) have further
discussed the pruning of the IMRT QA programme that has been possible as
194
the number of patients has increased. Miles et al (2003) summarized the median
times for performing various aspects of the planning and verification of IMRT and
showed that if the individual patient QA was removed, the times were comparable
between IMRT and conventional therapy. McNair (2002) has discussed the impact
on resources associated with the Royal Marsden NHS Foundation Trust phase-1
clinical trial of IMRT for prostate and pelvic node cancer. By March 2004, some
60 or so patients had been treated at the two geographical sites combined. McNair
et al (2004) discussed radiographers perspectives on the clinical implementation
but only through a qualitative survey of just this one centre after four months of
clinical work. De Langen et al (2003) have commented on the continued research
that is required once clinical IMRT is established.
IMRT has also been implemented at the Royal Marsden NHS Foundation
Trust (Fulham) for radiotherapy of the prostate with involved nodes. The clinical
technique was established on a Varian 2100 CD accelerator equipped with a
Varian Millennium 120-leaf MLC. Photons of 6 MV energy were used. Planning
was performed using the HELIOS system. IMRT was delivered in the dMLC
mode. The maximum leaf speed was set to 2.5 cm s1 . The MLC controller
monitors and records the position of the leaves every 50 ms. If the leaves are not
within a tolerance, then the beam is held off until they are. The HELIOS planning
system implements the technique of Spirou and Chui (1998). The resolution
is 2.5 mm along the leaf direction and a leaf-motion calculator converts the
fluence patterns to leaf patterns taking account of all the physics of transmission,
leaf speed, maximum leaf span, rounded leaf-ends (the so-called dosimetric leaf
separation [see also section 3.2.3]) and minimum leaf gaps etc. A leaf-position
tolerance of 2 mm was used. The leaf transmission was input to CADPLAN as
1.7% for 6 MV photons and 1.8% for 10 MV photons, these being averages of the
(variable) measured transmission at different depths. Measurements showed the
dosimetric offsets were 1.05 and 1.65 mm for 6 and 10 MV beams respectively,
using a measurement technique due to LoSasso et al (1998a, b). Planning
used five beams at angles of 180, 270, 325, 35 and 100 chosen after
planning experimentation. 6 MV was used as the higher 10 MV energy showed
no advantages. Planning skill involved sequential adjustments to the relative
importance given by the optimization to each organ and a gradual approach to
achieving the dose-volume constraints. Malden et al (2001) and Clark et al
(2002a) described these processes together with the verification of IMRT using
film measurements with the beams replanned on to a phantom. Measurements
were compared with plans for IMRT of the prostate with pelvic-node involvement.
This was done by exporting the plan on to a phantom comprising film, delivering
IMRT to the phantom and making comparisons between the scanned film and the
treatment-planning system (figure 5.6). Dose differences of up to no more than
2% were observed in regions of low dose gradient within the prostate and up to
3% in regions of low dose gradient within the nodes. Portal imaging with an
amorphous silicon imager showed agreement between measured fluence patterns
and calculated fluence patterns. The whole IMRT process required 23 hr per
Royal Marsden NHS Foundation Trust pelvic and other IMRT
195
Figure 5.6. An example of how a film measurement is used to check the accuracy of
delivery of a modulated beam. The upper panel is an overlay of isodose from film and
CADPLAN at 10 cm depth. The lower panel is the difference map of isodoses from film
and CADPLAN at 10 cm depth. (From Clark et al 2002.) (See website for colour version.)
patient compared with 13 hr for non-IMRT of a prostate only (a detailed table

breaks this down into tasks). With experience, the IMRT time was expected to
fall. Clark et al (2003c) has described this pruning of the QA programme for
IMRT at the Royal Marsden NHS Foundation Trust (Fulham).
Staffurth et al (2003a) have made a planning comparison of six patients
planned with CORVUS version 3.0, HELAX TMS version 6.0 and CADPLAN
196
version 6.3.5 with HELIOS. The CORVUS-planned delivery was simulated as a

dMLC delivery from an Elekta accelerator in service mode. The HELAX plan
was created as a segmented MLC plan from Elekta accelerators with Precise
Desktop version 3 and the CADPLAN was simulated as a dMLC technique from
a Varian 2100 CD accelerator. It was found that the PTV coverage was similar
across all three treatment-planning systems but that the number of MUs varied
dramatically and the shielding of OARs varied between the different planning
systems.
Staffurth et al (2002) have shown that IMRT of the prostate and involved
pelvic nodes has reduced the patient-reported acute GU toxicity compared to a
cohort treated with 3D CFRT to the prostate alone. Staffurth et al (2003b) have
shown through a phase-2 trial of pelvic-node irradiation in prostate cancer that
GU toxicity was significantly reduced; GI toxicity did not change. In another
centre, Bayouth et al (2003a) have similarly reported the ability to dose escalate
to the prostate whilst also treating the pelvic nodes with small GI toxicity.
Nutting and Dearnaley (2002) presented an overview of the complete IMRT
programme at the Royal Marsden NHS Foundation Trust. A planning study
had taken place using 30 patients with head-and-neck, pelvic and thoracic
tumours to compare conventional radiotherapy, 3D CFRT and inverse-planned
IMRT. The greatest improvements arose for thyroid carcinoma and pelvic-lymphnode treatment because these tumours have a concave PTV. For pelvic tumours
nine, seven or five equispaced coplanar IMRT fields gave similar benefits but
dose distributions deteriorated with three equally spaced fields. Conversely, for
oesophageal and head-and-neck tumours where structures with low radiation
tolerance were close to the PTV, it was important to use non-equispaced beams
and a special computerized algorithm was designed to calculate the best noncoplanar IMRT beam directions. Multi-field IMRT with only three to four beams
was then shown to be quite effective. To quality control the delivery technique,
treatment plans were delivered to a humanoid phantom and calculated doses were
compared with the doses measured with photographic film, TLD and BANG gel.
Adams et al (2001, 2003, 2004) have described the use of the Nordion
HELAX treatment-planning system for planning IMRT at the Royal Marsden
NHS Foundation Trust. Example applications are treating prostate cancer with
pelvic-node involvement, sparing the bladder and rectum. A planning study
has also been carried out to look at the complexity of treatment delivery for
maxillary sinus tumours attempting to significantly reduce dose to optic nerves
and ipsilateral parotid gland. Specific attention has been placed to minimizing
the number of field segments because the treatment was delivered with an Elekta
accelerator with (at that time) a large intersegment dead time. Plans were exported
from HELAX using the DICOM facilities and replanned on to a phantom for
experimental verification. TLD measurements have shown that, for a complete
five-field treatment, mean discrepancies were about 3%.
Comparison of IMRT with CFRT for complex shaped tumours
197
Figure 5.7. Solid geometry representation of tumour mass very close to OARs in the head.
Here a meningioma of the right sphenoid bone with retro-orbital extension is adjacent to
neighbouring brain stem, optic chiasm and lacrimal gland. (Reprinted from Pirzkall et al
2000 with copyright permission from Elsevier.)
5.4 Comparison of IMRT with conformal radiotherapy

(CFRT) for complex shaped tumours
Pirzkall et al (2000) have studied a comparison of IMRT and geometricallyshaped CFRT for nine patients extracted from the DKFZ Clinic during 1997. The
goal of the study was to assess the relative improvement that can be realized in
terms of planning quality as well as the impact of such an implementation on
a busy department. CFRT plans (those used to actually treat the patient) were
created using the VOXELPLAN 3D treatment-planning system using contour
data from correlated CT and MRI slices 3 mm thick with 3 mm separation.
For six of the targets, a manually driven MLC with a leaf width of 5 mm at
isocentre was used and for the other three a commercial Siemens MLC with
10 mm leaf widths was used. Planning limits were set such that preservation
of normal function was paramount because many of the lesions were benign
and patients were expected to live many decades post treatment. Preserving
normal structure was a clinical priority (figures 5.7 and 5.8). IMRT plans
were made with CORVUS 3.0 and segmentation was performed separately for
CORVUS and for VOXELPLAN implying no contour transfer between systems.
Hence, equivalence of segmentation was confirmed by evaluating the slice-byslice images and post-planning volumetrics statistics.
For each case, three IMRT plans were created: a rotational plan for serial
tomotherapy, a seven-field coplanar plan and non-coplanar plan. All plans
employed ten intensity levels. A range of evaluation metrics were used based
on the RTOG metrics for stereotactic treatments. These included target coverage
to desired dose, target coverage to CFRT dose, target conformality, target
heterogeneity, integral dose, MUs, treatment time and planning time.
198
Figure 5.8. Solid geometry representation of tumour mass very close to OARs in the
head. Here a chondrosarcoma of the skull base is adjacent to neighbouring brainstem,
optic chiasm and nerves, spinal cord. (Reprinted from Pirzkall et al 2000 with copyright
permission from Elsevier.)
The key outcome was that all three of the IMRT techniques produced
improvement over the CFRT plans in terms of conformality and coverage. They
were equivalent in terms of homogeneity, required a greater number of MUs,
delivered a greater integral dose and took less time to plan. There were essentially
no statistical differences among the three IMRT techniques. The most outstanding
observation was the improvement in the parameter target-coverage-to-CFRTdose, this being the target-volume-greater-than-the-CFRT-desired-dose divided
by the total target volume. For benign tumours, this parameter rose from about
0.72 to about 0.98 and for malignant tumours, it rose from 0.81 to about 0.96.
Trade-offs in IMRT included a delivery of higher integral dose outside the
target and an increase in MUs between two and eight times the number of MUs
per field compared to the CFRT plan. The authors acknowledged that they are not
the first to report the benefits of IMRT compared with CFRT nor to stereotactic
radiation therapy. However, they claimed that, in all their studies, IMRT was
found to be superior overall and this lends weight to the arguments against the
criticism of IMRT that it has yet to win its spurs. There is a lengthy discussion of
the results centring around what can be done with adjusting the various trade-offs
available to the IMRT planner.
Conversely, Vaarkamp et al (2003) have shown that some concave target
distributions can be created using five MSFs with up to four segments per beam.
The shielding of the OARs is not as great as from full IMRT but the plans are
considered clinically acceptable.
IMRT for whole-pelvic and gynaecological radiotherapy
199
Figure 5.9. Volume rendering of the clinical target volume for a typical patient receiving
whole-pelvic IMRT. The CTV contains the upper part of the vagina and parametrial tissues.
In women with an intact uterus, the entire uterus was included. The presacral region was
included to the bottom of the S3 vertebral body to ensure coverage of the presacral lymph
nodes and attachment of the uterosacral ligament. The common internal and external iliac
nodal regions were included to the level of the L4-5 interspace. (Reprinted from Roeske et
al (2000b) with copyright permission from Elsevier.)
5.5 IMRT for whole-pelvic and gynaecological radiotherapy

Gynaecological tumours present another opportunity for radiation therapy
improvement with IMRT. Small bowel and rectum are generally included in
the irradiated volume of conventional whole pelvic radiotherapy. Roeske et al
(2000b) have studied the ability of IMRT to reduce the volume of small bowel
irradiated in women with gynaecological malignancies who are receiving whole
pelvic radiotherapy (figure 5.9). Usually this disease is treated with a standard
four-field box with apertures shaped to the PTV in each beams-eye view, a
situation that leads to a significant amount of small bowel being unwantedly
irradiated as well as more bladder and rectum irradiation than desirable. Hence,
the main acute and chronic toxicity for these patients are small-bowel sequelae
including small-bowel obstruction, enteritis and diarrhoea and malabsorption of
vitamin B12, bile acids and lactose.
200
Roeske et al (2000b) studied ten consecutive women with cervical or

endometrial carcinoma, planning these patients for conventional radiotherapy
using the PLUNC 3D treatment-planning system. They then replanned the
patients using the CORVUS 3.0 inverse-planning system for nine fields arranged
coplanarly and delivering 6 MV photons. This latter choice was made following a
preliminary study to determine that, whilst the outcome of IMRT was dependent
on the number of fields, the improvement plateaued out after seven to nine
fields, consistent with the observations of others. They then showed comparative
plan data, comparative dose-volume histograms and comparative dose statistics,
extracted for all of these patients, which demonstrated the improved small-bowel
sparing of IMRT. A particularly interesting histogram shows the absolute volume
of the small bowel irradiated in each patient to a dose of 45 Gy or greater
indicating reductions often greater than a half between conventional radiotherapy
and IMRT. Similarly, tables showed improved sparing of dose to small bowel and
rectum. The price paid was a greater dose inhomogeneity in the PTV and normal
tissues but the hot spots were considered not to be of clinical significance because
their magnitude and volume were small.
The authors remind us that the Harvard Joint Center for Radiation Therapy
originally discussed computer-controlled radiation therapy in the context of
pelvic-node irradiation in the 1970s, but, given the available technology at the
time, the research was not taken forward to the clinic. The important new work in
the current study provides planning proof that IMRT is worthwhile and it remains
to be seen whether the benefits translate to reduced treatment sequelae. The
patients are being followed closely to answer that question. This work took place
at the University of Chicago.
Roeske et al (2000a) have shown that, by using nine intensity-modulated
fields to the pelvis, cervical cancer can be treated without the requirement for
intracavitary brachytherapy. With IMRT, only 10% of the rectal volume received
the prescription dose whereas this value was 30% when brachytherapy was
combined with conventional whole-pelvic radiotherapy.
Low et al (2000a, 2002b) have developed a technique known as applicatorguided IMRT for gynaecological cancer. An intracavitary applicator substitute
was placed in the patients vagina and uterus and 3D images were acquired using
MR. An IMRT plan was then computed using the CORVUS planning system
to conform to the applicator substitute. Hence, when the dose distribution was
aligned with the applicator substitute, it was also aligned with the tumour and
critical organs. This technique is expected to replace brachytherapy at this centre.
Wahab et al (2004a) have given further details of the technique known as
applicator-guided IMRT to deliver IMRT to the cervix. The aim of this study
was to define the target volume using positron emission tomography (PET) and
then use IMRT instead of the conventional radioisotope high-dose-rate applicator.
It was shown that this leads to a more uniform homogeneous dose to the target
avoiding the higher low-dose regions that are typical of brachytherapy.
Head-and-neck IMRT
201
Portelance et al (2001) have shown that IMRT can improve the

conformality of dose to the cervix with improved rectal and small-bowel sparing.
Conventional radiotherapy was planned using the CMS FOCUS system and
either two or four fields. IMRT was planned with the NOMOS CORVUS
system using either four, seven or nine fields. The volume of small bowel
receiving the prescribed dose fell from 35.58% (conventional two-fields) and
34.24% (conventional four-fields) to 11.01% (four IMRT fields), 15.05% (seven
IMRT fields) and 13.56% (nine IMRT fields). The volume of rectum receiving
more than the prescribed dose fell from 84.01% (conventional two-fields) and
IMRT fields) and 3.34% (nine IMRT fields). The volume of bladder receiving
more than the prescribed dose fell from 92.89% (conventional two-fields) and
IMRT fields) and 26.91% (nine IMRT fields). The differences between the IMRT
techniques were deemed insignificant, four fields being as good as nine. It was
determined that dose escalation to the cervix would be possible with the IMRT
techniques. The study was based on the step-and-shoot technique.
Hong et al (2002) have assessed the feasibility of inverse planning for wholeabdomen IMRT. They created a PTV comprising the entire peritoneal cavity with
a margin. Five 15 MV intensity-modulated beams were created at gantry angles
180, 105, 35 , 325 and 255 and IMRT optimization was designed to spare
kidneys and bones. Significant dose reduction to the bones and improved PTV
coverage were achieved and, because of the very large width of the beams, to
minimize field match errors adjacent subfields overlapped by at least 2 cm with
intensity feathering in the overlap region.
Vuong et al (2002) delivered IMRT for unresectable rectal or rectosigmoid
cancer and showed that dose escalation could be achieved with acceptable
toxicity.
5.6 Head-and-neck IMRT

So far in this chapter we have tried to separate the statements about the prediction
of benefit from those about its actual observation. Clinical IMRT of the head-andneck was reviewed by Guerrero Urbano and Nutting (2004a,b). Eisbruch (2002)
has emphasized that there is a large difference between the expectation of benefit
from IMRT and the demonstration of this. He points to IMRT for head-and-neck
treatments in which it has recently been conclusively demonstrated that reduced
xerostomia arises (Lee et al 2002a). It was observed that xerostomia gradually
reduced, post treatment, to almost zero after two years. Head-and-neck cancer
presents a very real clinical test for IMRT since there are so many potentially noninvolved tissues adjacent to tumour including major and minor salivary glands,
mandible, pharyngeal musculature, inner and middle ears, temporomandibular
joints, temporal brain lobes and optic pathways.
202
Munter et al (2003) have described the treatment technique for stereotactic

IMRT of head-and-neck tumours at DKFZ Heidelberg and its inverse treatment
planning. Forty-eight patients with carcinoma of the head-and-neck region
were treated between 1999-2002. Planning was performed with the KONRAD
treatment-planning system and treatment was through interpretation by the
Bortfeld technique to segments delivered by a Siemens accelerator. The technique
is very successful with only four patients having xerostomia greater than grade
1. The authors described the now well-known QA techniques implemented at
Heidelberg to determine absolute and relative dosimetry (see section 3.13.9).
There is a current European head-and-neck IMRT-versus-conventionaltherapy protocol and trial (Guerrero Urbano and Nutting 2004a). It is intended
to recruit 160 patients. The participating centres are The Netherlands Cancer
Institute, University of Ghent, Christie Hospital, Manchester and the Royal
Marsden NHS Foundation Trust. The Marsden is coordinating the trial. IMRT
can only be routinely implemented by UK NHS hospitals after phase-3 clinical
trials have given, what is called, level-one evidence of benefit (Bentzen 2004).
5.6.1 Thyroid IMRT

Parker et al (2000) have shown how IMRT can improve the coverage of the PTV
for radiotherapy treatment of the thyroid. Four- or five-field directions are used
with a few (four to five) MLC subfields for each.
Happersett et al (2000) have also developed IMRT for the treatment of
thyroid cancer. It was shown that, compared with a conventional technique,
IMRT can significantly increase PTV coverage while maintaining similar or better
normal structure doses. The normal structures concerned were the lung and spinal
cord.
Clark et al (2002b) have clinically implemented dynamic IMRT for patients
with carcinoma of the thyroid and cervical nodes (figure 5.10). The prescription
doses were 60 Gy to the thyroid PTV and 50 Gy to the nodal PTV respectively.
Five-field IMRT plans were designed with the aim of reducing unnecessary
anterior shoulder irradiation and also including the use of asymmetric fields for
improving spinal cord sparing. It was found that the average maximum dose
received by the cord could be reduced to below 47 Gy. The dose range in
the thyroid PTV was 4.2 Gy and the range was 10.6 Gy in the nodal PTV
with all beams covering all of the targets. The dose range in the thyroid PTV
was 6.2 Gy and in the nodal PTV was 11.2 Gy, respectively, with asymmetric
posterior oblique beams. Comparison of measured doses with calculated doses
gave differences of 1% in the thyroid PTV and just 2% in nodal PTV. The
conclusion was that IMRT offered improved target homogeneity for patients with
thyroid carcinoma as well as reducing the dose to the spinal cord.
Head-and-neck IMRT
203
Figure 5.10. Two CT slices showing the nodes of a patient treated for cancer of the thyroid
and larynx. The right-hand picture is of the nodes superior to the primary volume and the
left-hand one is of the nodes inferior of the primary volume. Notice the highly irregular
volume that it is required to irradiate conformally necessitating IMRT. (Courtesy of Dr
Catharine Clark.) (See website for colour version.)
5.6.2 Nasopharynx IMRT

Xia et al (2000a) have, for one treatment-planning case only, that of a
locally advanced nasopharyngeal carcinoma, compared four treatment plans
(figure 5.11). These were a conventional treatment, a non-modulated conformal
treatment, an IMRT plan generated using CORVUS and deliverable with MSFs
and an IMRT plan created using PEACOCKPLAN for the MIMiC. The MSF plan
had five gantry angles chosen manually with ten non-zero intensity levels for each
beam. The 3D conformal plan had eight gantry angles based on the beams-eye
view. In practice, the plan was delivered using partial transmission blocks and the
study was just for technique comparison.
QA was carried out by delivering the plan to a phantom comprising a stack
of solid-water slabs with the approximate dimensions of a human head. Relative
dose distributions were verified using Kodak XV-film and point-dose distributions
were checked with an ionization chamber. Conformality of the plans was ranked
such that the best plan was the PEACOCKPLAN, followed by the CORVUS plan
204
Figure 5.11. Axial, coronal and sagittal displays of GTV in red and CTV in orange for a
selected nasopharyngeal case. The complex re-entrant surface and shape of the volumes to
be treated is evident and calls for IMRT. (Reprinted from Xia et al (2000a) with copyright
permission from Elsevier.) (See website for colour version.)
followed by the 3D conformal plan, followed by the conventional bilateral plan.

Xia et al (2000a) also assessed the dependence of the plans on set-up uncertainty
by a technique of shifting the isocentre relative to the plan.
Hsiung et al (2002) have compared IMRT with conventional 3D CFRT
for boost or salvage treatment of nasopharyngeal carcinoma (NPC). Fourteen
NPC patients planned with the ADAC PINNACLE planning system in Taiwan
were transferred to the Memorial Sloan-Kettering Cancer Center planning system
where new IMRT plans were produced using five or seven radiation fields. Plans
were then compared and it was shown that the IMRT plans were superior.
Chu and Suh (2000) have delivered IMRT for NPC using compensating
filters. Van Nimwegen et al (2001) have also developed IMRT for the
nasopharynx. A planning study comprised 20 patients and a five-field technique
was used. Lu et al (2004) have shown how IMRT can be used for re-treatment of
NPCs after failed conventional radiotherapy.
5.6.3 Oropharynx IMRT
Maes et al (2000) have developed a class solution using five IMBs for treating
oropharyngeal cancer whilst sparing the parotid gland. The technique was based
on the HELIOS inverse-planning system from Varian.
Head-and-neck IMRT
205
Van Asselen et al (2001a, 2004) have worked towards a class solution

for oropharyngeal cancer. It was found that it was possible to systematically
determine a standard set of dose constraints which resulted in a homogeneous
dose distribution to the target and acceptable critical organ dose. Seven equiangular beams were enough to obtain an acceptable dose distribution using
between five and ten intensity levels in the sequencing procedure.
5.6.4 Oropharynx and nasopharynx IMRT
Van Dieren et al (2000) have shown that intensity modulation for the treament
of midline tumours of the head-and-neck can spare parotid gland irradiation.
Fifteen patients were studied, each with two modulated portals. Sparing of the
submandibular glands was less pronounced for tumours of the oropharynx. For
primary tumours of the larynx, all salivary glands could be spared.
The IMBs were generated 2.5 mm apart in the longitudinal direction
corresponding to a compensator with this resolution. Four adjacent profiles were
averaged to produce the equivalent modulation generatable with an MLC with
1 cm wide leaves. It was found that the clinical outcome was barely affected by
this process (see also section 3.9.3).
5.6.5 Larynx IMRT
Levendag et al (2000) have performed a multi-institutional evaluation of IMRT.
Twenty-two patients with cancer of the larynx without lymph-node metastasis
were treated by 3D CFRT to the primary site and to the neck. A CT scan dataset
of one (typical) clinical case out of this series was then subsequently selected and
with the pre-contoured target and normal tissues was re-planned independently
in Amsterdam, Ghent, Brussels and Rotterdam using IMRT techniques. The data
provided interesting comparisons between the techniques of the four centres.
Wu et al (2001e) and Mohan et al (2001b) have developed a simultaneous
integrated-boost technique for advanced head-and-neck squamous cell carcinoma
using the dMLC technique. Similar work, reported by Dogan et al (2003),
claimed that the main advantage was greater conformality and the use of a single
plan throughout treatment. Conversely, Popple et al (2003) have developed
planning software that plans both the unboosted phase and the boost phase
simultaneously for sequential delivery. Neither of the two phases necessarily
respects the treatment plan constraints alone but the sum of the two plans always
respects the constraints.
Clark et al (2002c, 2003a, 2004) have developed IMRT for six patients
treated for larynx carcinoma in a two-phase treatment. They concluded
improved target homogeneity was achieved with reduced dose to the spinal cord
(figure 5.12). Clark et al (2003d) explain the rationale for the use of five, rather
than more, fields for some head-and-neck cancers, and also how the use of cutouts in the immobilization mask can reduce skin dose. Lee and Xia (2003)
206
a)
b)
c)
Figure 5.12. Dose distribution for phase-1 treatment of the larynx and nodes to 50 Gy
using the IMRT technique. The transverse cross sections are (a) through the central larynx
showing the larynx PTV, the nodal PTVs and the spinal cord, (b) the superior nodes
showing the nodal PTVs and the spinal cord and (c) the inferior nodes showing the nodal
PTVs and the spinal cord. (From Clark et al 2004.) (See website for colour version.)
conversely consider that the lack of skin reaction at the Institute of Cancer
Research/Royal Marsden Hospital (ICR/RMH) may have more to do with the
use of a lower (46 Gy) prescribed dose to nodal regions compared with 60 Gy at
University of California at San Francisco (UCSF).
5.6.6 Evidence for parotid sparing
Vineberg et al (2000) have shown that the use of optimized beamlet IMRT inside
the UMPLAN treatment-planning system in Michigan has led to bilateral parotid
Head-and-neck IMRT
207
sparing. The contralateral parotid dose could be reduced to a mean of 26 Gy

while the ipsilateral parotid ranged between 26 and 35 Gy. These results could
be improved by increasing the number of beams from four to nine leading to a
reduction in parotid mean dose of up to 34 Gy. However, inclusion of non-axial
beams did not improve parotid sparing.
Bragg et al (2001) have also shown that IMRT of the parotid gland can spare
the contralateral parotid provided suitable beam directions are chosen.
For head-and-neck planning at The Netherlands Cancer Institute, three to
eleven coplanar fields are used, each with two-to-four segments, the goal being
to spare parotid glands and to irradiate a concave PTV. Segments were predetermined; the dose from each segment found and then an iterative looping was
used to compute segment weight. It was found that the volume of the parotid
raised to more than 25 Gy was reduced from 92% to 48% and the conformity
index (the irradiated volume greater than or equal to 95% divided by the PTV)
fell from 2.5 to 1.6. It was thus concluded that IMRT was advantageous and also
improved PTV homogeneity.
Braaksma et al (2002) have shown that some IMRT techniques can increase
dose to the minor salivary glands thus tending to offset sparing of the major
parotid glands. Braaksma et al (2003) have evaluated the parotid sparing
of 3D CFRT in a prospective study in node-negative cancer of the larynx.
Twenty-six patients were treated and dose distributions of the major salivary
glands were correlated with objective stimulated whole saliva flow and subjective
(questionnaire: visual analogue scale) salivary-gland function. It was found that
the whole-saliva flow reached its nadir six months post-radiotherapy and then
improved significantly such that at two years post-treatment measurements were
48% of the pre-treatment values. However, the class solutions for 3D CFRT
salivary-gland-sparing technique were inadequate for preserving fully salivary
gland function and it was determined through a planning study that further
reduction of xerostomia could be achieved using an IMRT technique focused at
sparing major and minor salivary glands. This technique was developed through
a cooperative effort of four centres (Daniel den Hoed Cancer Centre, Rotterdam;
University of Ghent, Belgium; The Netherlands Cancer Institute, Amsterdam;
and the University of Michigan, Ann Arbor. These Centres all used different
inverse-planning techniques. Following this, all 26 patients receiving 3D CFRT
were re-planned with IMRT, showing that the mean parotid gland dose reduced
to 23 Gy compared with 28.9 Gy for 3D CFRT. It was also determined that IMRT
by the dMLC technique gave very similar results to IMRT performed with the
step-and-shoot technique in five patients. The dose-volume histogram findings
being superimposable. This study provided important quantitative evidence for
the efficacy of IMRT for treating head-and-neck cancer (figure 5.13).
Bel et al (2002) have produced treatment plans in which the position of the
head-and-neck tumour was sampled randomly from known distributions of headand-neck tumour position and IMRT plans re-computed. These were compared
to the plan created with just a single unmoved PTV. The comparison showed
208
Figure 5.13. The upper panel shows the objective relationship between mean dose (Gy)
in the parotid glands (x-axis) and residual salivary gland function (expressed by the
whole-saliva flow rate, % of baseline, y-axis) at 1-year post-irradiation. The lower panel
shows the correlation between mean dose in the parotids and subjective salivary-gland
function (expressed by visual analogue scores (VAS), y-axis). The white squares/dots
represent patients in whom a correlation between mean dose and salivary gland function
can be depicted (three and six patients); black squares/dots represent patients in whom
dose distribution in the parotid glands and outcome on xerostomia do not correlate (8 and
11 patients). Full bold horizontal and vertical lines represent thresholds (Reprinted from
Braaksma et al (2003) with copyright permission from Elsevier.)
no change with respect to the dose to the clinical target volume and PTV but
considerable changes to the NTCP of the parotid glands.
Mazurier et al (2002) have implemented IMRT using a HELAX system and
a Siemens accelerator. They found that predicted and measured dose distributions
agreed to better than 5% and applied the technique to reduce parotid irradiation
to avoid xerostomia.
Head-and-neck IMRT
209
Figure 5.14. This shows the seven co-planar equidistant intensity-modulated fields used to
treat a 70-year-old female with extensive sphenoid wing meningioma treated with IMRT.
The patient was treated with IMRT for a recurrence after two operations (4 years and
10 months before IMRT). The plan was created with the KONRAD treatment-planning
system from Heidelberg. (Reprinted from Pirzkall et al 2003 with copyright permission
from Elsevier.)
5.6.7 Meningioma IMRT

Debus et al (2000) have irradiated 17 patients with large skull-based meningiomas
using IMRT in a phase-2 study. The clinical requirement was to treat the complex
shape of the meningiomas. Dynamic MLC treatments were delivered with a
Siemens Primus accelerator with an integrated MLC and a step-and-shoot IMRT
delivery technique. Treatment was quality-assured using film dosimetry and an
ionization chamber. Transient acute-treatment side effects were monitored.
Pirzkall et al (2003) reported on a follow-up of these patients using MRI
and clinical examination showing no radiation-induced peritumoural oedema, no
increase in tumour size and no new onset of neurologic deficits. IMRT was
considered the treatment of choice (figures 5.14 and 5.15).
Uy et al (2002) have reported on 40 patients treated for intracranial
meningioma (excluding optic nerve sheath meniongiomas) using the NOMOS
210
Figure 5.15. The intensity pattern of one IMRT treatment portal: (A) as calculated by
the treatment-planning system KONRAD; (B) as acquired during quality assurance and
dose verification using a BeamviewPlus portal-imaging system from Siemens Oncology
Systems. (Reprinted from Pirzkall et al 2003 with copyright permission from Elsevier.)
PEACOCK CKPLAN system between 1994 and 1999. The cumulative five-year
local control was 93%.
Some clinical evidence for the benefit of IMRT for treating primary optic
nerve sheath meningioma has been presented by Woo et al (2002).
5.6.8 Paediatric medulloblastoma IMRT
Papanikolaou et al (2002) have shown that IMRT can reduce the dose to the
kidneys and the thyroid gland in the treatment of paediatric medulloblastoma and,
as a result, reduce the incidence of clinical and subclinical hypothyroidism.
5.6.9 Ethmoid cancer IMRT
Wang et al (2002b) have performed a comparative study of IMRT versus 3D
radiotherapy for treating ethmoid cancer. They found that IMRT gave superior
sparing particularly to the lens of the eye. For the Varian 80-leaf MLC, they
stated that the physical limitations of the MLC led to a degradation of the IMRT
plan.
5.6.10 Other studies reported
Cozzi et al (2001a) have compared three different treatment techniques: 3D
conformal therapy, intensity-modulated photons and intensity-modulated protons.
Breast IMRT
211
Five patients were studied and plans were computed for all treatment modalities.
The patients presented with advanced head-and-neck tumours and OARs included
the spinal cord and the parotid glands. It was concluded that protons provided
significantly the best improved dose homogeneity and also the best sparing of
spinal cord. The intensity-modulated photon plans were superior to the nonmodulated photon plans but inferior to proton plans. The same results were found
for the parotid glands. Cozzi et al (2004) further amplified on the benefits of
IMRT for treating head-and-neck tumours.
Cozzi et al (2001d) further compared six different treatment techniques for
advanced head-and-neck cancers. These techniques included conventional mixed
electronphoton beams, 3D conformal photon beams with five fields, intensitymodulated photon beams with either nine or five equally-spaced fields and three
proton fields either with passive scattering or spot scanning. Plans were evaluated
using dose-volume histogram data and predictions of equivalent uniform dose. It
was found that all the conventional techniques were inferior to other techniques
using IMRT or protons and that, in turn, the proton distributions were superior to
those using photon IMRT. Zurlo (2002) comments that IMRT to targets in large
parts of the body delivers a higher integral dose than protons to normal tissues.
He also calls for international collaboration on trials of proton therapy.
Morgan (2001) used the HELAX TPS to create custom compensation
for head-and-neck treatments. Using the Elekta Desktop delivery system and
accelerator, the accuracy was verified to better than 2% with isodoses better than
3 mm.
Pawlicki et al (2004) have estimated that the lens dose can be an appreciable
fraction of the treatment dose in some head-and-neck IMRT treatments. They
compared predictions from CORVUS and Monte Carlo calculations. It is
important, therefore, to avoid irradiating through the eyes.
Burnet (2003) showed that the use of rotation with a central-axis block could
be an alternative to IMRT for spine tumours.
Erridge et al (2003) used a simple set of (four) top-up fields added to an
85% open field to perform electronic compensation for head-and-neck tumours to
improve PTV uniformity.
5.7 Breast IMRT

Guerrero Urbano and Nutting (2004b) have reviewed clinical IMRT of the breast.
5.7.1 Breast IMRT at William Beaumont Hospital
Kestin et al (2000a, b) described the William Beaumont Hospital, Royal Oak,
Michigan technique for performing IMRT of the breast (see also Wavelength
2001c,e, 2002b). The technique begins by calculating the dose distribution
from open tangential fields using tangential beams with correct account of tissue
inhomogeneity. Separate MLC segments were then constructed to conform to
212
the beams-eye-view projections of the 3D isodose surfaces in 5% increments

ranging from 100120% (figure 5.16). Generally, this leads to the generation
of between six and eight segments. The beamweights of these segments
were then optimized using an inverse-planning technique and by placing 100
reference points uniformly distributed throughout the breast PTV. A special
script programme for the ADAC PINNACLE treatment-planning system was
then created to optimize the weights of the individual segments by conducting
a downhill Simplex search on the 100 randomly placed points in the PTV. It was
found that, in general, this technique reduced the number of field segments to
between two to four segments per beam direction. Any beam segment that was
required to deliver less than four MUs was deleted. A second IMRT plan was
then created by the same technique but including lung blocking. It was found
that the largest component of dose to the breast was delivered from the open
fields (as expected) and that IMRT considerably improved dose homogeneity to
the breast compared with the use of standard wedges. It was found that the most
dramatic outcome was the reduction from around 10% of the volume receiving
110% of the planning dose down to just 0.1% of the volume when IMRT was
constructed with segmented beams with lung blocking. There was also a small
decrease in the maximum dose and mean dose. The modest number of segments
kept treatment delivery to less than 10 min and the dose homogeneity to the breast
improvement was expected to correlate with improved cosmetic outcome. A
simplified account of the work at the William Beaumont Hospital appeared in
Wavelength (2000a) emphasizing the technique of creating MSFs by following
isodose lines and reweighting the components. This is a specific form of DAO
(Shepard et al 2003a) based on the use of isodose lines rather than on projections
of target and at-risk structures.
This group considers that this is much more practical than using custom
physical compensators but the technique at the William Beaumont Hospital does
rely on the use of available high quality CT data. This distinguishes the technique
from the early techniques at the Royal Marsden NHS Foundation Trust, which did
not use CT data.
Vicini et al (2002) have recently reported that the treatment-planning system
has been upgraded to permit dose optimization over the entire breast using dosevolume objectives and constraints. Also, whereas formally segments with very
low numbers of MUs were rejected, segments with as few as 2 MUs are now
routinely used.
By April 2001, 281 patients had been treated with this technique. The
number of segments varied depending on the breast size, with more segments
being required for patients with larger breasts. Typically between five and eight
segments were required and the median treatment time was less than 10 min.
The authors presented the dose-volume outcomes and reported that no skin
telengiectasias nor fibrosis nor persistent breast pain had been noted in any of the
patients. The cosmetic results at 12 months had been rated as excellent or good in
99% of the patients. The authors claimed that widespread implementation of this
Breast IMRT
213
Figure 5.16. Beams-eye view of an MLC segment with a block corresponding to the 105%
open-field isodose surface. The image demonstrates a 2D central-axis reconstruction of the
isodose surfaces. Dark blue = 9094%; light blue = 9599%; lavender = 100104%;
red = 105109%; yellow = 110114%; white = 115119%. The segments defined in this
way then have their weights optimized. (Reprinted from Kestin et al 2000a with copyright
214
technology could be achieved with minimal imposition on clinical resources and

constraints. Specifically, they claimed that their technique does not significantly
increase the treatment-planning nor dose-delivery times and they cite rival breast
IMRT techniques as being less efficient in these respects.
An editorial (Potters et al 2003b) considered the paper of Vicini et al (2002).
Strom (2002) postulated that the most important impact of the change in planning
radiotherapy of the breast using this IMRT technique is on the physician, namely
how the physician thinks about the breast and the operative bed as targets. For
example, the protocol gives a strict method to tailor the tangential fields to the
individual patients anatomy. Being CT-based, it also allows a boost volume as
well as the primary breast tangential fields to be constructed. It is postulated that it
largely eliminates differences in dose-prescribing conventions, which are caused
by a variety of traditional prescription points because the 100% minimum breast
dose must be specified in a tissue region and it also challenges the understanding
of what exactly is the clinical target volume.
The treatment technique was verified by delivering the modulated radiation
to anthropomorphic phantoms comprising slices of tissue-equivalent material
containing, between slices, radiographic film and it was concluded that
measurements were within 3% of treatment plan calculations. The vast majority
of the dose (median of 78%) was delivered by the open segments. Finally it was
postulated that the technique would probably be less susceptible to the effects of
respiratory motion, a hypothesis that is currently being evaluated in this clinic.
5.7.2 Other reports of techniques using small top-up fields

Donovan et al (2002b) have made a dose-histogram analysis of 300 patients, half
randomized to conventional tangential irradiation and half to IMRT. In standard
wedged treatment doses greater than 105% were largely in upper and lower thirds
of the breast. Ninety-six percent of the patients had doses greater than 105% in
the upper third and 70% had doses greater than 105% in the lower third. Only
4% of patients receiving IMRT had doses greater than 105% in either third. The
volume of breast receiving greater than 105% dose decreased by a mean of 10.7%
when treatment changed from conventional to IMRT. A goal of the randomized
trial was to correlate this improved dosimetry with patient-reported accounts of
the location of pain and breast tenderness (figure 5.17).
Van Asselen et al (2001b) from Utrecht have also developed a technique to
improve the dose uniformity to the breast by using IMRT (figure 5.18). The basis
of the technique was to deliver 88% of the dose using two open fields covering
the whole treated volume and then the remaining 12% is delivered in four MLCshaped segments. They found that this decreased the dose inhomogeneity in the
PTV from 9% for the conventional technique to 7.6% for the IMRT technique.
At the same time, the mean lung dose was reduced for the IMRT technique by
approximately 10% compared with the conventional technique.
Breast IMRT
215
Figure 5.17. An example differential dose-volume histogram plot for a standard and for an
IMRT plan. Dotted lines are the standard plan, full line is the IMRT plan. (From Donovan
et al 2002b.)
Figure 5.18. The four MLC settings for each intensity level in IMRT of the breast as
deduced for the equivalent pathlength map. The first segment includes the areas 14, the
second segment the areas 24, the third segment the areas 34 and the fourth segment area
4. (Reprinted from van Asselen et al (2001b) with copyright permission from Elsevier.)
In addition to the goal to improve dose homogeneity, the intention was

to maintain a vertical match plane between the tangential fields and the
supraclavicular and axial fields without using a breast board. The IMRT technique
used was a step-and-shoot technique using an Elekta accelerator. Wedges could
not be used to optimize the dose distribution because in such an accelerator the
wedge direction is perpendicular to the leaf-movement direction. Hence, a MSF
IMRT technique had to be developed. For this, 3D CT data for the patient had
to be available. Three-dimensional dose distributions were calculated using the
Nucletron PLATO RTS version 2.2 planning system and the energy of the beams
was 6 MV. The first attempt made was to improve the dose distribution in the
central plane of the breast. This was done by calculating equivalent pathlengths
taken from ray tracing from the focus of the beam through the CT data set
216
projected on to a plane perpendicular to the beam axes. The translation of such

an equivalent-pathlength map into optimal MLC field settings and intensity levels
was a complex problem. However, this map was used to obtain the MLC settings.
In general, for each of five patients, just four such MLC settings were used. The
resulting dose plans were analysed in terms of the volume of tissue with dose
lying between particular isodose values. The dose distributions displayed by
isodose contours in the central plane showed an improved dose distribution for
the IMRT technique compared with the conventional technique. The ipsilateral
lung dose decreased with a mean value of 11.8% compared with the conventional
technique. Prior to the use of intensity modulation, it was observed that the patient
with the largest breast had the worst dose homogeneity. For such patients, the use
of more intensity levels or optimized weightings of the segments might result in a
further increase in dose homogeneity. It was felt that the ipislateral lung would be
spared radiation pneumonitis via the technique. Apart from the use of CT data, the
technique has much in common with that at the Royal Marsden NHS Foundation
Trust and has similar conclusions.
Remouchamps et al (2000) have evaluated whether segmented tissue
compensation was superior to the use of physical wedges when treating largebreasted patients. They concluded that, in all respects, it was. Segmented tissue
compensation significantly decreased the overdosage in the inframammary fold
and axillary region. Ten patients were studied in detail.
Cozzi et al (2001c) have shown that IMRT of the intact breast improves on
non-IMRT treatments especially with more than two fields.
Li et al (2004c) have also developed a breast IMRT technique based on the
use of CT data. The conventional tangential field directions were selected (with
the option to make small changes) and the bixel intensities were set firstly to
optimize the dose at the midplane and then to spare dose to normal structures,
lung and heart. Final dose calculations were made by Monte Carlo techniques.
As with earlier studies by Donovan et al (2002b), improved sparing of dose to
normal structures was demonstrated as well as a more homogeneous dose to the
target volume.
5.7.3 EPID-based techniques for breast IMRT
Heijmen (2000) has shown that EPIDs can be used to design beam profiles for
IMRT (e.g. for treating breast cancer) and that the EPIDs can also be used for
verifying the thickness profiles of the compensators so produced (see also Evans
et al [2000] figure 5.19). The pre-treatment dosimetric verification of beam
profiles for the dMLC technique has also been investigated and clinically applied.
Methods have also been developed for realtime verification of leaf positions
during dMLC treatments. Megavoltage CT has also been developed based on
the use of EPIDs.
There have been many techniques suggested for IMRT of the breast. Chui
et al (2002) have proposed another. In this technique, all that is required is the
Breast IMRT
217
Figure 5.19. Superposition of a patient image from the Royal Marsden NHS Foundation
Trust in-house EPID and the four fields to be delivered from this beam direction. A left
medial image is shown. This is overlaid with the fields needed (a) for flash method 1 and
(b) for flash method 2. The labels inside the images indicate the field numbers. (Reprinted
from Evans et al (2000) with copyright permission from Elsevier.)
external outline of the breast in the direction of the tangential beams. The beams
are then divided into 2 mm 2 mm bixels and the intensity of each bixel is
inversely proportional to the dose delivered to the centre of the line of sight of the
line intersecting the breast from that bixel. The aim then is to produce the same
dose to all the central positions of such lines. It is important to note that these
lines do not necessarily lie in a single plane. The intensity profile is then extended
as is usual to produce a skin flash region. Chui et al (2002) compared this
simplified IMRT technique with a volume-based IMRT technique and a wedgedpair technique for 15 patients with left-sided breast cancer and showed that the
dose homogeneity to the breast from the simple technique was considerably
better than that from the wedged pair and was similar to the outcome from a
full volumetric technique. The main time saving on treatment planning with this
simplified technique relative to the volumetric technique comes from eliminating
most of the contour delineation that is necessary for the latter. In some ways, this
method has similarities with those at the Royal Marsden NHS Foundation Trust
from Donovan et al (2000) who also attempted to equalize either the mean dose or
the maximum dose in intersecting trajectories. However, this technique required
218
the use of a portal-imaging device and other volumetric techniques such as that by
Landau et al (2001) required the use of CT data and full inverse planning. Chui
et al (2002) averaged five adjacent profiles of width 2 mm to produce a profile
that could be delivered with a 10-mm-wide MLC leaf. They also showed that
the simplified IMRT technique provided better sparing to the lung in the highdose region than the conventional method and lower dose to the ipsilateral lung
and contralateral breast. The simplified IMRT technique is practical for largescale implementation in a busy clinic without requiring significant increase of
resources.
5.7.4 Modified wedge technique
Luo et al (2000) have produced a breast radiation technique comprising a
tangential set-up with a thin medial wedge and thicker lateral wedge. The
advantage of this is a reduction of between 525% in the dose to the contralateral
breast, a matter of particular importance when irradiating young women.
5.7.5 Breast IMRT in combination with use of respiration gating
Dyke et al (2001) have developed an IMRT technique for left-sided breast
cancer. This technique is combined with a respiratory-gating device developed
by Varian Medical Systems which synchronizes the radiation treatment with the
patients breathing pattern (see also sections 6.10.3 and 6.10.7.1). At the time
of simulation, two CT scans are performed in immediate succession, the first
with the patient holding their breath on exhale and the second on inhale. During
treatment planning, the radiation oncologist will then determine from the dosevolume histograms whether to gate the treatments on exhale or inhale. Once
this is established, three or four MLC fields are designed to reduce the dose
inhomogeneities within the breast.
5.7.6 Comparison of IMRT delivery techniques for the breast
Partridge et al (2001a) have compared several IMRT techniques for delivering the
top-up modulations needed to create the intensity-modulated fields for tangential
breast radiation therapy. The first was the MSF technique in which, for the
Elekta accelerator operating at 6 MV, 2N control points are required for N equalfluence fluence increments. The second was the dMLC technique, operated in
close-in form, in which (N + 1) control points are required for the same N
equal-fluence increments (see figure 5.20). For completeness, a sliding-window
delivery was also carried out but with the proviso that this delivered the whole
fluence including the baseline constant fluence to which the increments are added
in the other two techniques. The basis of all these deliveries is that the ideal
fluence modulations are computed first, independent of machine constraints, and
then these are interpreted. Measurements were made both with and without
Breast IMRT
219
Figure 5.20. Schematic diagram illustrating the numbers of control points generated
by dividing a continuous intensity modulationhere with just one peakinto (a) three
discrete MSFs and (b) a series of three steps for a dynamic delivery. (From Partridge et al
2001a.)
the presence of an additional wedge fluence modulation. The number of control

points was varied from 4, 10 to 15.
Measurements were made with film with appropriate build-up and were
repeated three times to assess statistical error. Absolute dosimetric measurements
were identical for four different delivery options (15-control-point dynamic with
and without wedge and 14-increment MSF with and without wedge.) The MU
efficiencies of both MSF and dynamic delivery techniques were very similar
and decreased slightly as the number of control points increased. In terms of
total delivery time, the dynamic delivery techniques were faster than the MSF
techniques, the speed decreasing as the number of control points increased due
to the several machine checks and deadtime at each control point. The use of the
wedge increased the delivery time in all corresponding situations partly due to the
need to rotate the head for the wedged portion of the delivery.
Measurements of dose profiles showed very good agreement between MSF,
dMLC and ideal dose for 14 fluence increments (15 control points). Use of the
wedge only changed the dosimetry in the flash region. The effect of reducing the
number of control points (fluence increments) was measured. It was determined
that at least 10 are needed for fields without wedges and four for fields with
wedges. Close-in also well approximated dynamic delivery. Thus, the use of
220
Figure 5.21. Calculated NTCP values for late cardiac mortality for three planning
techniques (grey bar = rectangular fields; white open bar = conformal technique; black bar
= IMRT. (Reprinted from Hurkmans et al (2002) with copyright permission from Elsevier.)
the wedge whilst minimizing the number of top-up fields required, an advantage,
conversely led to a disadvantageous increase in treatment time.
Kerambrun et al (2003) have compared IMRT of the breast with tangential
fields with and without wedges, showing improvement in target homogeneity and
lung sparing. The IMRT was verified using a sliced phantom with 25 films and
accuracy obtained was 3% in the low-dose region and 3-mm high-dose isodose
lines.
5.7.7 Reduced complications observed following breast IMRT
Many studies have already shown IMRT of the breast leads to better dose
homogeneity. Hurkmans et al (2002) set out to demonstrate that IMRT also
reduced radiation pneumonitis and increased dose sparing of the heart so reducing
the NTCP for late cardiac mortality (figure 5.21). Importantly, they included in
the study the computation of the effects of including geometrical field shaping
only as a midway stage between the use of rectangular tangential fields and full
IMRT. They found that the NTCP for radiation pneumonitis was 0.3%, 0.4% and
0.5% for IMRT, conformal shaping and rectangular shaping. The NTCP for late
cardiac mortality was 2.0%, 4.0% and 5.9%, respectively, for these techniques.
Seventeen randomly selected left-breast treatments were included in the study.
Non-IMRT planning was on UMPLAN and IMRT planning used KONRAD.
Breast IMRT
221
There was concern that the (CT-based) IMRT technique would lead to a
much larger clinical burden. It was shown that the cardac NTCP correlated
well through a second-order polynomial with the maximum heart distance, a
geometrical parameter which indicated the amount of heart in the field. This
was used to estimate which patients would need IMRT. Cardiac NTCP was
estimated using the data from Gagliardi et al (1996). This is a planning study
and observation of actual clinical outcomes are awaited.
Cho et al (2001a) have shown that IMRT of the breast decreases the dose
to the heart and both lungs compared to wide tangential-field irradiation. Cho et
al (2002b, 2004) have shown that a simple class solution for IMRT is possible
for IMRT of the left breast and can lead to significant cardiac sparing compared
with non-IMRT approaches. Cho et al (2002a) have shown that a breast IMRT
technique gives comparable results to that of an oblique electron plus photon
technique when both breast and internal mammary chain are included in the
irradiation. A wide-field photon tangent technique gave worse results. All inverse
planning, performed on KONRAD, gave segments which were re-optimized using
UMPLAN with a better dose model. Cho (2002) has discussed IMRT class
solutions for treating breast cancer reducing the risk of cardiac complications for
left-sided breast cancer patients. It was argued that even small relative gains in
survival and recurrence-free outcome would translate to large absolute gains.
At NKI Amsterdam, work on IMRT of breast cancer in which the planning
system KONRAD was used has shown that IMRT could reduce the normal
tissue complication probability of cardiac mortality to 2% whereas it was 4% for
conformal therapy and 5.9% for simple rectangular fields. A similar diminution
of NTCP for radiation pneumonitis was also observed. This was based on the
NTCP model of Gagliardi for heart, showing that NTCP of heart increases with
maximum heart distance, being about 6% when the maximum heart distance is
3 cm.
Thilmann et al (2002) have developed IMRT for breast cancer when internal
mammary and supraclavicular lymph nodes were included in the treatment
volume at DKFZ, Heidelberg. It was found that the volume of the ipsilateral
lung irradiated with a dose higher than 20 Gy was reduced from 26% to 13%
compared with conventional treatment. For patients with left-sided tumours, the
heart volume with a dose higher than 20 Gy was reduced from 11% to less than
1% and the technique resulted in improved homogeneity in the target volume. In
conclusion, there were substantially improved dose distributions.
Aznar et al (2001) showed that the use of additional segmented fields in
the treatment of breast cancer led to a higher and narrower differential dosevolume histogram reflecting a homogeneous dose distribution. This effect was
not observed in plans created using either compensators or mixed beam energy.
A clinical trial is being set-up to evaluate the impact of this improvement on
cosmetic outcome and quality of life in patients.
222
5.7.8 Combination of IMRT with charged-particle irradiation

Fogliata et al (2002) have compared a series of different photon plans with and
without IMRT to the results obtainable using protons for breast cancer treatments.
Not surprisingly the proton treatments were always superior in terms of both the
dosimetry of the PTV and OAR sparing. However, a three-field IMRT plan also
gave very good conformality.
Li et al (2000a) have shown that the dose to normal structures near the
breast can be considerably reduced by combining a single electron field with
four intensity-modulated photon fields. This was far superior to the use of two
tangential fields and comparable to, if not better than, the use of nine IMRT fields
without electrons.
5.8 Bladder IMRT at the Christie Hospital

Budgell et al (2001b) have described the custom-compensated four-field
treatment of carcinoma of the bladder that has been used at the Christie Hospital
NHS Trust, Manchester, as a simple test site for the introduction of IMRT.
Modulations were converted to MLC leaf-and-jaw settings for dynamic delivery
on an Elekta linear accelerator. A full dose calculation was carried out and
a test run of the delivery was performed prior to treatment. Absolute dose
measurements were made in water or solid-water phantoms. Treatments were
verified by in-vivo diode measurements and also by electronic portal imaging. At
the time of writing, seven patients had been treated using the dMLC technique.
This report brought together the results and observations of a number of
earlier technical papers from this group but it was the first to report in detail on the
full scope of the treatment technique and its specific clinical implementation. The
Christie philosophy is that clinical implementation of IMRT must be approached
with the utmost care and therefore the intention was to tackle simple problems
prior to full implementation of IMRT. Logue (2000) at the Christie has described
how IMRT in the short term is considerably more demanding of resources and
more expensive than conventional therapy. It is therefore important to define the
clinical needs for IMRT and to define potential benefits. The selection of the
appropriate delivery technique needs to include consideration of the complexity,
deliverability and efficiency of the technique. It is important to understand that
only through attention to all of the stages of the radiation therapy procedure
through planning to delivery can the full potential of IMRT be achieved. To
this end, although patients received 20 fractions over four weeks, in the initial
implementation only four of the fractions were replaced with compensated IMRT
fractions, two in the second and two in the third weeks of treatment. The
planning process was completely automated and intensity-modulated profiles
from each gantry angle were computed to deliver a homogeneous dose within
a defined target volume in the isocentric plane normal to the beam axis. The
leaf-synchronization technique of Budgell et al (1998) was used to ensure that
Bladder IMRT at the Christie Hospital
223
all leaves finished simultaneously and never exceeded their maximum speed.
Following this, a full dose calculation was performed based on the calculated
leaf trajectories and the use of a macro-pencil-beam algorithm. This takes full
account of leaf-and-jaw motions and of the transmissions through leaves and back
up jaws. Both head and phantom scatter were modelled at each control point. It
was possible to iteratively correct the intensity modulations to recalculate the leaf
settings to deliver the altered modulations. Subsequently the dose was calculated
for the whole 3D dose volume.
Prior to treating the patient, absolute dose measurements at the isocentre of
each beam were made using an ionization chamber in a rectangular water or solidwater phantom and the measured and calculated doses were required to be within
2% overall for treatment to commence.
Treatment was delivered using an Elekta accelerator operating in dynamic
mode at 100 MU min1 . The leaf speed was initially restricted to 5 mm s1 .
The positions of the leaves was checked 12.5 times per second and a beam-finish
facility was available. Additionally, diodes were used for dosimetric verification
and a portal imager for real-time geometric verification of leaf positions according
to a technique previously described by James et al (2000). Budgell et al (2001b)
showed a typical compensating intensity modulation for a lateral beam for a
dMLC bladder treatment showing the increased intensity required at the beam
edges to compensate for lack of scatter at the beam edge. It was found that
the technique was effective in reducing the high-dose areas found within static
plans (figure 5.22). An issue of some concern in all dMLC treatments is the MU
inefficiency and it was observed that the inefficiency factor (the ratio of MUs
required to peak MUs) lay in the range 1.371.75 comparing favourably, with for
instance, the 20 wedge factor of 1.51 for the same machine with the same photon
energy.
Clinical experience at the Christie began in April 1999 and, up to the end
of November 1999, seven patients had been treated. The documented additional
time per patient required for planning and delivery using that research system
was between 7 and 8 hr. Some of this was due to the necessity to change between
clinical and service mode and back again and also to manually change the gantry
angle between beams. However, the in-vivo dosimetry showed that all beams
measured during the pre-treatment dummy run lay within 3% of the calculations.
It was determined that a leaf-position accuracy of 1.5 mm was sufficient and that
this was achieved by the standard quality-control protocol. It was observed that
the Elekta MLC was very precise in dynamic delivery and the overall technique
took less time resources than for the production of manual compensators. The
group proposed to investigate similar techniques for the breast and for headand-neck cancer where much higher levels of compensation are required and
they concluded that, whilst IMRT of the bladder is not strictly necessary, the
experience paves the way for the introduction of more complex and clinically
beneficial IMRT techniques. A simplified account of the work at the Christie
Hospital appeared in Wavelength (2000a) emphasizing improved homogeneity of
224
Figure 5.22. A comparison of treatment plans with and without dMLC compensation: (a)
isocentric CT slice using static MLC fields (no IMRT); (b) isocentric CT slice using dMLC
IMRT technique; (c) most superior CT slice using static MLC fields; (d) most superior CT
slice using the dMLC IMRT technique. IMRT improves the target coverage and reduces
dose gradients across the target. (Reprinted from Budgell et al (2001b) with copyright
dose to the bladder. Hounsell et al (2001) have also described further the Christie
Hospital technique for IMRT.
McBain et al (2003), at a different centre, used MRI of the pelvis to assess
the time-dependent motion and volume of the bladder. Serial MR images were
viewed in cine loops. Patients had a characteristic volume-versus-time curve
which was largely reproducible and linear. Based on this, they developed POLO
(Predictive Organ Localization) a planning technique to take account of bladder
motion. There have actually been very few reports on clinical IMRT of the
bladder.
5.9 Lung cancer IMRT

Guerrero Urbano and Nutting (2004a) have reviewed the limited clinical IMRT
of the lung. Dirkx and Heijmen (2000a) have extended their technique to take
account of the lateral transport of radiation. They compared modulated dose
distributions in lung materials with standard treatments. They investigated the
technique for the improvement of the treatment of lung cancer. Compared with
Scalp IMRT
225
the treatment-planning study, the superior and inferior boost fields, used in the
beam intensity modulation (BIM) technique for penumbra enhancement, had
to have a higher weight and be longer in order to compensate for the lateral
transport of electrons in lung tissue. The technique of using reduced-length fields
with penumbra enhancement was then robust to this phenomenon. They also
showed that calculations made with the CADPLAN system matched closely the
measurements taken in model phantoms simulating the treatment of lung cancer
(figure 5.23).
Yorke (2001) has compared IMRT using small numbers (49) of beams with
manually optimized plans using static conformal beams for non-small-cell lung
cancer. It was concluded that, independent of the model of lung toxicity that was
employed (Lyman or parallel models), IMRT plans were consistently superior for
target coverage and normal-tissue protection.
Koelbl et al (2002) have compared CFRT and IMRT for ten lung patients.
They found that the dose to the ipsilateral lung was significantly reduced with
IMRT but dose to the contralateral lung increased within the lower dose levels,
probably without clinical significance.
Phillips (2002) has applied inverse planning using projection onto convex
sets to improve the radiotherapy of lung cancer.
Sethi et al (2001b) have built a phantom to simulate mediastinal irradiation
with lung inserts. They then created step-and-shoot and tomographic IMRT plans
at 6, 10 and 18 MV for this phantom and made measurements with and without
inhomogeneity corrections. It was found that the plan without inhomogeneity
corrections delivered overdose to the PTV. However, this overdose decreased with
increasing energy.
Conversely, and more controversially, Stevens (2003) argues that it is too
early to use IMRT for lung cancer with mobile targets. Superb gated and breathhold imaging studies have shown the vulnerability of target definition to tumour
motion. A study by Liu et al (2004) in the same group showed that, in principle,
it is possible to use IMRT of the lung to reduce the percentage of lung volume
receiving more than 20 Gy as well as the integral lung dose. IMRT also spared
dose to oesophagus, heart and spinal cord (Murshed et al 2004).
5.10 Scalp IMRT

Locke et al (2002) have compared conventional treatment for merkel cell
carcinoma of the total scalp with IMRT. They concluded that the IMRT plan
provided a more homogeneous dose to the target volume but the critical structure
doses were uniformly higher than for the conventional treatment plan and so, for
the time being, IMRT is ruled out.
226
Figure 5.23. (a) Measured isodose distributions in lung-equivalent material at a depth

of 12.5 cm for treatment of a PTV (shaded area) with the 10 MV photon beam using the
standard field (broken lines) and a 1.4 cm shorter field combined with BIM (full lines). The
isodose levels depicted are the 20%, 50%, 90% and 95% levels. (b) A comparison between
the measured isodose distribution for the BIM treatment (full lines) and the distribution
predicted by the CADPLAN treatment-planning system (broken lines). The 20%, 50%,
90%, 95%, 100% and 105% isodose levels are shown. (c) Measured dose profiles in
the superiorinferior direction along the central axis for the standard field (broken line)
and BIM (full line) and corresponding dose profiles predicted by the treatment-planning
systsem (dots and squares, respectively). (Reprinted from Dirkx and Heijmen (2000a) with
copyright permission from Elsevier.)
Other clinical IMRT reportsvarious tumour sites
227
5.11 Other clinical IMRT reportsvarious tumour sites

James et al (2002a, 2003) have introduced IMRT within a (so-called) small
non-academic hospital for eight patients who have been treated with five- or
seven-field IMRT to sites within the head-and-neck, pancreas and lung. For
head-and-neck tumours, IMRT allowed the construction of irregularly-shaped
target volumes avoiding OARs. In the treatment of large inoperable pancreatic
cancer, IMRT reduced toxicity by sparing normal small bowel and other sensitive
structures and, by using IMRT for paraspinal lung tumours, the high-dose region
could be shaped to avoid an unacceptable dose to the cord. The IMRT programme
will continue at Suffolk Oncology Centre, Ipswich, with the introduction of
further clinical sites. Crosbie et al (2002) described the QA of IMRT at this
Centre. They found that the calculated cumulative dose at the plan isocentre
was consistently greater than the corresponding measured cumulative dose in
a perspex block phantom by about 5%, an issue they are continuing to work
on. Secondly, the dose from each treatment field was measured separately in a
solid-water phantom and compared with the calculated doses. The measurements
and the calculations differed by approximately 4.5%. When calculated
fluences, plotted as isodose contours, were compared with measured fluences,
the agreement was within 3%. James et al (2002b) found that electronic
compensation leads to a better dose homogeneity within the breast and this was
borne out by phantom measurements in which measured and calculated doses
agreed to within 3%. Poynter (2003) also commented that the largest increased
workload in developing IMRT at the Suffolk Oncology Centre has been increased
time for treatment planning and for QA measurements. At March 2003, seven
head-and-neck cases (two neck, two thyroid, two paraspinal and one parotid
sparing), six lung cases and two GI cases had been treated. CFRT and IMRT plans
were both developed for each patient and patient selection for IMRT was made
if the IMRT dose distribution was superior. James et al (2003) gave an update
on clinical IMRT at the Suffolk Oncology Centre. By the time of this UKRO2
conference (April 2003), 21 patients had been treated (ten head-and-neck cases,
nine lung cases and two pancreas cases). The main observation was an increase
in the requirement for radiological contouring and about three hours per patient
of QA.
Hawliczek et al (2000) have used the CORVUS inverse-planning system
with the PRIMEVIEW/SIMTEC software at a MEVATRON KD2 Siemens
accelerator at the Institute for Radiation Oncology in Vienna. Eleven patients
with various indications had been treated and the technical installation dosimetry
and integration into clinical service was all performed within two months. They
concluded that, using turn-key IMRT systems, the technique was ready for routine
use under specific clinical conditions. Schmidt et al (2000) have shown that plugand-plan IMRT was possible in this hospital in Vienna. Schmidt et al (2001)
are using a Siemens step-and-shoot and a Varian sliding-field and step-and-shoot
IMRT system for routine IMRT treatment. They have particularly emphasized the
228
importance of understanding the scatter dose in patients consequent on the use of

IMRT.
Lopez Torrecilla (2001) has emphasized the steps of introducing IMRT
into the clinic. The pathologies which would benefit from IMRT were studied
and the importance of understanding the accuracy of patient repositioning was
recognized. In the first phase, a similar dose was used as used in CFRT to acquire
confidence in the new technique prior to dose escalation. It was emphasized
that different planning systems using different inverse-planning algorithms use
their constraints and beam weights differently and so a measure of trial and error
must be undertaken jointly by the physicist and radiation oncologist before full
understanding is obtained for any given system.
Mazurier et al (2001a) are using the HELAX treatment-planning system
coupled to a Siemens PRIMUS accelerator for IMRT. They have studied a prostate
case and a head-and-neck case and studied the whole chain of IMRT planning
and development. After regular calibration of the MLC, it was found that the
leaf-positional accuracy was better than 1 mm, that the monitor-chamber linearity
was about 3% for 1 MU and that the treatment-planning system had a tendency
to overestimate the dose in the penumbra area. A comparison between calculated
and measured dose distributions was considered acceptable.
Rosello et al (2000) have used the HELAX treatment-planning system with
a PRIMUS-Siemens linear accelerator to deliver IMRT in step-and-shoot mode.
They discussed the chain of QA tests required to implement the technique.
Henrys et al (2003) used an IMRT technique with just a few segments to
improve the PTV coverage for colorectal cancer.
Marcie et al (2002) have implemented IMRT using the HELAX system and
a Siemens Primus accelerator.
Aletti et al (2002) have used the CADPLAN inverse-planning system with
Varian 23 EX Clinac delivery to treat ten patients in the dynamic mode.
Scielzo et al (2002) have described a standard work-up for IMRT using the
Millennium-120 MLC.
5.12 Summary
At the time of writing, the situation with regard to clinical implementation is very
non-uniform. There are some centres who have developed their own techniques
or adopted commercially available ones and who have worked up clinical IMRT
for large patient numbers and sometimes (though not often) as part of a clinical
trial. These have published full papers at some length and it is possible to obtain
a clear picture of their achievements.
With respect to some of the earliest clinical IMRT at Memorial Sloan
Kettering Cancer Center (prostate) and Royal Marsden NHS Trust (breast),
genuine clinical outcome data are available supporting the thesis that IMRT is
needed and advantageous. Some evidence of reduced xerostomia exists too.
Summary
229
However, most implementation is on the basis of belief, based on planning studies,

that clinical outcome justifies the investment.
Regarding organ site, much attention has been given to IMRT of the prostate,
breast, head-and-neck and bladder. Lung IMRT, gynaecological IMRT and other
organ IMRT is more patchy. Perhaps not surprisingly, many of the clinical reports
are at present just a set of conference abstracts that make it hard to get a clear and
full picture and certainly leads to a quite patchily constructed review.
Chapter 6
3D planning for CFRT and IMRT:
Developments in imaging for planning and
for assisting therapy
6.1 Challenges to IMRT and inverse planning

In this chapter, we shall move away from the techniques to deliver IMRT and its
clinical application and now consider some of the bedrocks on which the science
of IMRT is based. Specifically these include the application of high-quality 3D
medical imaging to assist the planning of IMRT and its verification. Also the
use of medical imaging to assist coping with inter- and intrafraction movement
will be considered. All of these concepts can be grouped under the generic title
image-guided (intensity-modulated) radiation therapy (IGRT or IG(IM)RT). This
chapter also groups together reviews of modern inverse-planning and forwardplanning techniques.
In particular, it may be important to create inhomogeneous dose distributions
within the PTV depending on the proximity of parts of the PTV to OARs
or because there is specific information about the non-uniformity of tumour
cells and function. IMRT thus requires us to consider how to obtain the
required information and the best use of the ability to design these dosimetric
inhomogeneities. Pelizzari (2003) has stressed the importance of investigating
the dependence of IMRT on new positional information determined through
SPECT, PET and MRI. Using those imaging modalities which have higher spatial
resolution, it may be necessary to redesign the IMRT delivery, e.g. to use finer
leaves in multileaf collimation to produce IMRT with steeper dose gradients. It is
also important to consider the effects of patient movement and the complexity of
delivery which in turn affects the resource requirements. IMRT also challenges
us to consider new ways of evaluating complex treatment plans. The development
of IMRT has challenged much conventional thinking in radiotherapy and requires
us to reconsider the whole chain of radiotherapeutic processes.
Determination of the GTV, CTV and PTV
231
6.2 Determination of the GTV, CTV and PTV; the influence

of 3D medical imaging
In this section, we discuss the determination of the GTV, CTV and PTV. The
growing role of 3D imaging modalities is highlighted. The issue of accounting
specifically for tissue movement will be reviewed in section 6.10.
6.2.1 General comments on inhomogeneous dose to the PTV and
image-guided planning
Tepper (2000) has emphasized that we are on the brink of a new era in which
functional imaging will play a much larger part in treatment planning than in the
past. A major task will be to integrate the images in order to effectively use them.
At the same time, there has been progress in understanding how volumes can be
determined from (more traditionally used) x-ray CT data.
Ling (2001a) has also emphasized the importance of using biological
images for planning IMRT. These images should include predictive assays
of radiobiological characteristics. Ling (2001b) argues that the planning of
IMRT should be driven using images of metabolic, biochemical, physiological
functional and molecular information. Images that give information about tumour
hypoxia that in turn influences radiosensitivity and treatment outcome can be
regarded as radiobiological images. Hypoxia is considered one of the main
limiters to radiation response (Chapman et al 2003).
Price (2002) has also discussed the introduction of functional and molecular
imaging for volume definition particularly in relation to IMRT and CFRT. It was
argued that functional imaging is likely to improve the definition of the GTV
but not necessarily the CTV due to inadequate spatial resolution. It was also
questioned whether the identification of hot spots for dose boost using functional
imaging is desirable. Functional imaging can also be used to determine critical
areas to avoid. Kilovoltage CT at the time of treatment will enable correlation
of the target volume to be treated with the functional images determined pretreatment (see section 6.11.2).
6.2.2 Interobserver variability in target-volume definition
In the UK, an important clinical trial, known as the Medical Research Council RT01 trial, is assessing the effect of dose escalation in the prostate from 64 to 72 Gy.
This has involved 13 participating centres. An important part of the trial was to
ensure that the centres all operated to high and consistent standards. Because the
radiotherapy process is a chain, starting with the outlining of targets and normal
structures, a study was carried out to determine the inter-clinician variability of
outlining (Seddon et al 2000). Three clinical datasets of CT images on film were
sent to the 13 participating centres together with instructions and also a test set
already completed to demonstrate the process. The outlines were returned to the
232
3D planning for CFRT and IMRT
coordinating centre (the Royal Marsden NHS Foundation Trust) and these were
analysed as well as being transferred by hand and without modification to the
CADPLAN treatment-planning system. Beams-eye views of the PTV were then
created and analysed for their inter-clinician variability. It was also shown, by
re-analysing one set of data by one clinician with a long time interval between,
that the transfer process was acceptably accurate. The following was observed:
(i) there was remarkably little inter-observer variability in the GTV central-slice
identification;
(ii) there was good consistency in outlining the inferior bladder and the superior
bladder;
(iii) the definition of the superior rectum was very variable;
(iv) the prostatic apex was very hard to define;
(v) the definition of the base of the seminal vesicles presented problems.
As a result, at the Royal Marsden NHS Foundation Trust, it is now common
practice to include the use of MR images in the work-up towards the planning
of prostate cancer. These observations correspond well with the experience of
clinicians and it is not really surprising that these differences have been observed.
Moeckli et al (2002) have conducted a study investigating how oncologists
in 11 Swiss centres delineated target volumes according to ICRU-50
recommendations for a prostate and a head-and-neck case. The results were
astonishingly varied between the centres with great discrepancies indicating no
clear consensus on interpreting diagnostic images.
6.2.3 Margin definition
Dobbs (2000) and Graham (2000) have emphasized the importance of
understanding the determination of the PTV and the associated geometrical
uncertainties. Complementing this, MacKay and Amer (2000) have modelled
the effects of movement of the target volume with respect to CFRT in terms of the
change in TCP and normal tissue complication probability.
McKenzie et al (2000) have shown that the margin which should be added
to the CTV is given by
(6.1)
w = 2.5 + ( p )
where is the standard deviation of the systematic set-up error on the CTV, p
is the standard deviation of the penumbra function for a single beam and is the
standard deviation of the random movement errors (figure 6.1). Van Herk et al
(2000) previously showed that = 1.64 is the appropriate value for a single beam
irradiation to ensure that the whole of the PTV lies within the 95% dose envelope
and McKenzie et al (2000) extended this to consider the effect of superposing a
number of beams to create the high-dose volume. The value of was given in
these circumstances for up to six beams. The derivation is somewhat empirical for
idealized conditions but this is commented upon extensively. McKenzie (2003a)
233
Figure 6.1. The doctors delineation of the CTV results in a systematic error . In 90%
of cases, the mean position of the CTV will be entirely encompassed by drawing a margin
2.5 wide around the delineated CTV (From McKenzie et al (2000)).
has used this methodology to derive a typical margin for a prostate CTV. As
well as applying a margin to the CTV, margins should also be applied to OARs
(McKenzie et al 2002) via
w = 1.3 0.5.
(6.2)
This margin is smaller than the CTV-to-PTV margin because movement of the
OAR away from the beam is helpful whereas any movement of the CTV is
unwanted. McKenzie (2003b) described the new British Institute of Radiology
Report on geometric uncertainties in radiotherapy.
Visser et al (2000) have shown how systematic and random variations in
patient positioning can be included into radiation therapy treatment planning.
In general, the margin should be given by approximately twice the standard
deviation of the systematic variations plus 0.7 times the standard deviation of the
random variations. Other workers have shown similar but not identical numerical
relationships.
McKenzie (2000) has provided a mathematical argument which shows that
the correct way to include the effects of varying tumour position due to breathing
(for example in the upper thorax) is to add the full amplitude of the breathingrelated motion to the width of the CTV and only then add the sum of the
quadrature errors due to set-up, machine geometry and beam penumbra. The rule
of thumb is that the CTV must be kept inside the full breathing-accommodated
volume and this is the opposite of expecting that uncertainties, including the
breathing motion, should all be added in quadrature.
234
Often the data for these standard deviations are not available and so resort is
made to adding margins determined through experience. Deshpande and Poynter
(2000) have shown that increasing the superiorinferior margin of the prostate to
10 mm instead of the usual isotropic 8 mm used in the Medical Research Council
RTO1 dose-escalation trial for CFRT of the prostate leads to an increased TCP
achieved at the expense of only a small increase in low-grade rectal toxicity.
Jackson (2002) has shown that increased rectal-shielding and posterior-PTV
margin sizes of about 0.6 cm reduce rectal complication rates in IMRT and
prostate cancer.
Kirby and Shentall (2001) have shown that IMRT is more sensitive to
patient movement or set-up inaccuracy than conventional therapy and advise an
additional 0.2 cm margin to give the IMRT treatment the same CTV underdose
with movement as would be obtained for a non-IMRT treatment with the smaller
margin.
Caudrelier et al (2000) evaluated the use of fuzzy logic for volume
determination in MRI and demonstrated that it improved accuracy and
reproducibility. The fuzzy logic method to determine the volumes of interest
(VOI) asks observers to give a confidence to the contours being created. The
observer specifies a maximum contour outside which a structure definitely does
not lie and a minimum contour inside which a structure definitely does lie. The
region in between is the fuzzy region and is parametrized by a sigmoid acceptance
function (Caudrelier et al 2003). This contrasts sharply with the classical method
of obtaining a volume by scanning the contour and multiplying by the slice
thickness (over all slices). A number of phantoms containing VOI structures of
different shapes and sizes were imaged by MR and the VOIs determined both
ways. The classical method was often erroneous by up to 70% whereas the fuzzy
logic method was accurate to 2%. The former was not robust with respect to slice
thickness whereas robustness is a feature of fuzzy logic techniques.
6.2.4 Use of magnetic resonance for treatment planning
There has been wide reporting of the use of MR to assist the definition of target
volumes in treatment planning.
6.2.4.1 Distortion correction
Tanner et al (2000) have presented details of the Royal Marsden NHS Foundation
Trusts technique for removing (or at least correcting for as far as possible)
distortions in pelvic MR images which will be used for planning radiotherapy
treatment. They observed that the distortions were primarily of two types: those
due to the MR equipment and those due to the presence of the patient. The
technique presented referred to the former. These were a consequence of the
nonlinearities in the field gradients and inherent inhomogeneities in the main
(a)
235
(b)
Figure 6.2. (a) Linearity test object mounted within the reference frame phantom, (b)
schematic exploded view of components shown in (a): (i) reference frame, (ii) linearity
test object, (iii) couch insert. (From Tanner et al 2000.)
applied magnetic field B0 . Uncorrected images displayed geometric distortions

of up to 16 mm, unacceptable for planning radiotherapy.
The patient was positioned on a flat couch insert to simulate the treatment
position. A map of system-generated distortions was created by designing a
special linearity test object to measure the difference between the expected (true)
positions and the apparent positions (figure 6.2). It was very important to
determine a specific set of MRI parameters which remained the same both for
the measurements of the test object and for those of the patient. The test object
comprised a series of tubes which could be filled with contrast material and which
then showed as either points or lines in the images. Additionally, a ring of marker
spots was located at the field periphery. The linearity test object was designed
in such a way that there were enough tubes for the purpose but that they were
adequately spaced so as not to interfere with the measurement.
Software was created to automatically locate the marker positions. This was
the centre of gravity of the pixels indicating the markers. Then a distortioncorrected image was created by assigning to each pixel in the new image an
interpolated value from the distorted image. Tanner et al (2000) developed
consistent QA procedures for the process. In particular, before each new patient,
the images of the external ring of points, distortion-corrected, was compared with
the reference distortion image and this determined the subsequent actions, either
acceptance of the correction or a request for a recalibration.
Experiments showed that by applying consistent imaging procedures, stable
system distortions could be maintained over many weeks with mean changes of
less than 1 mm in the measured positions of the rods. It was also shown that the
system corrections (mean shift) were highly dependent on the MRI parameters
236
(a)
(b)
Figure 6.3. A pelvic image (a) before and (b) after correction for system distortion. (From
Tanner et al 2000.)
demanding these stay consistently the same. For this reason, when an MR scanner
is upgraded or replaced, these measurements and calibrations must be repeated.
For patient imaging, it was shown that the corrections were not so stable
because of the effect of the changing size of the patient (figure 6.3). The largest
patient (actually volunteer) led to errors of some 3 mm. It was left to future work
to characterize these patient-related distortions.
Wu et al (2004) have shown a method to de-distort images of the prostate to
obtain a map of citrate to aid the planning of intraprostatic lesions.
6.2.4.2 Use of contrast agents

Barentsz (2000) has shown how contrast-enhanced MRI can assist with planning
radiotherapy of the pelvis. This capitalizes on the improved detection and staging
that is consequent on the use of contrast-enhanced MR. Gadolinium-enhanced
MRI can show the tumour extension. New more-specific lymph-node contrast
agents can also show the pelvic metastatic lymph-node spread. This may replace
lymphography.
Beavis and Liney (2002a, b) and Beavis (2003a) have used the
paramagnetic agent gadolinium-diethylenetriaminepentaacetic acid (DTPA) to
produce dynamic contrast enhancement of solid tumours visible with MR. This
has enabled them to define boost volumes for simultaneous delivery with IMRT.
237
6.2.4.3 Planning based on MR images alone

Fransson et al (2000a) have developed treatment planning based on MR images
alone. A 0.2 T Siemens Open Vision MR system generated images which
were segmented by ANALYSE into tissue types. Backfrieder et al (2000)
segmented MR data for converting the images into Hounsfield units for radiation
therapy treatment planning. The segmented MR images were compared with coregistered CT images and accurate image characterization was achieved using
manually selected training sets. The tissues characterized were air, bone, bonemarrow, adipose tissue, muscle, brain, cerebrospinal fluid and tumour. The
segmented tissue regions were then assigned appropriate CT Hounsfield values
and the converted MR data were input into a HELAX TMS treatment-planning
system for calculation of dose distributions. Results were compared to the
corresponding CT-based dose distributions and found to differ by no more than
1%. They concluded that MR can provide the basis for treatment planning. It
was not clear whether the same conclusion would have been obtained if the MR
contours would have been different from the CT contours as is sometimes the case
(see section 6.2.4.4).
Lee et al (2003) have shown that T1 -weighted MR images can be
segmented and bulk-assigned electron densities can be allocated to each segment,
distinguishing air, water and bone (T1 is the MR longitudinal relaxation time).
Plans made on such images gave dose distributions differing by no more than 2%
from those created using x-ray CT images.
Chen et al (2003) have also shown that MRI alone can be used to plan IMRT
for the prostate.
6.2.4.4 Coregistered CT and MR planning
Technique
Toner et al (2000) have evaluated the accuracy of using external MR- and CTcompatible fiducial markers for image fusion in stereotactic treatment planning.
They placed six markers external to a Rando head phantom for both imaging
modalities and three markers internally which were used for evaluation. Image
fusion was then performed using the BrainLAB software to match the centroids
of structures. Spatial errors for the registration were then computed from the
displacements of the centroids of the internal markers for three separate fusions
using either three, four or six external markers. The error in the registration had
a maximum value of 1.6 mm. It was concluded that multimodality markers as
the basis of a fusion tool were sufficiently accurate for most stereotactic planning
cases.
Fransson et al (2000b) have registered MR and CT data using both the
BrainLAB 3-point technique and the HELAX eight-anatomical-landmark-point
technique. Five anatomical regions were outlined on the MR images and
transferred to the coregistered CT data and the registration outcome was evaluated
238
based on a visual inspection of these five regions on the CT images. It was found
that the two techniques were rated similarly but that the BrainLAB technique took
longer. However, this was compensated by a more user-friendly environment and
also the additional manual adjustment of the registration which was allowed by
the BrainLAB technique.
Van Herk et al (2003) showed how chamfer matching and mutual
information techniques can generate merged CT/MRI data for delineating target
structure. Maps of the urological structures (like world maps) were constructed
to show the differences in delineation of the structures in the two modalities.
Matching algorithms are also a good way to quantitate organ motion.
CT and MR definition of the prostate
Most comparative studies have concluded that MR is a very useful adjunct to CT
for prostate definition. However, reports do not lead to a consistent statement
concerning which volume is greater and whether rectal sparing improves or not
with the use of MR images for planning.
A study by Parker et al (2003) of comparative CT and MR definition of the
prostate showed:
(i)
(ii)
(iii)
(iv)
MR and CT volumes were almost identical;

interobserver variation of GTV definition was lower for MR than for CT;
MR-assisted definition of the prostatic apex and
gold seeds provided a useful registration tool (figure 6.4).
Freedman et al (2001) have shown that the use of MRI can indicate the
prostate better than CT. MRI data were used for creating IMRT treatment plans
and it was shown that, although the MRI prostate volume was larger compared
with that from CT, the MRI-based treatment plan did not result in increased dose
to the bladder or rectum. Conversely, Lebesque et al (2001b) found that MRI
generated much smaller target volumes for the prostate, some 30% smaller, just
as was found for head-and-neck. Also Steenbakkers et al (2002) showed that, for
IMRT of the prostate, the dose delivered to the rectum was significantly reduced if
the treatment planning was based on an MR delineation of the prostate compared
to a CT delineation. Vanregemorter et al (2002) have reported the volume of the
prostate viewable using CT and MR prior to CFRT.
De Meerleer et al (2002b) have used IMRT to deliver a higher dose to an
intraprostatic lesion than the dose given to the rest of the prostate. To do this, they
used a high-resolution turbo T2 weighted MR image allowing delineation of the
intraprostatic tumour.
CT and MR definition of head-and-neck tumours
Gregoire (2002) indicated that functional MRI and PET imaging could be used to
define sub-GTVs, for example highly hypoxic or proliferative areas in which an
239
Figure 6.4. (a) An axial CT image through mid-prostate. The fiducial marker is well
visualized but the prostate gland is not easily distinguished from the surrounding soft
tissues. (b) The corresponding FSE sequence axial MR image from the same patient.
The fiducial marker appears as a signal void (arrowed). The prostate outline is clearly
demarcated from surrounding tissues. In addition, a dominant intraprostatic lesion is seen
in the right peripheral zone. (Reprinted from Parker et al (2003) with copyright permission
from Elsevier.)
240
extra boost dose could be sculpted using IMRT, for example, the sparing part of
the salivary gland with high secretory activity. A key issue is that images need to
be obtained under a very similar immobilization conditions.
Fenton and Lynch (2002) have shown that the use of MRI data, combined and
registered with CT data, changed the volume of brain excluded from the tumour
volume by more than 50% in six out of eight of the cases studied.
Chytka et al (2000) have used parameters extracted from CT and MR
to distinguish between low- and high-grade glioma without performing a
stereobiopsy. Using a statistical model, the discrimination was found to be 83%
accurate when applied to 214 patients.
Aoyama et al (2001) have shown how registering MR and CT data can
reduce the inter-observer variability in determining the GTV. They also showed
that the MR-GTV was significantly lower than the CT-GTV for cerebellum/brainstem lesions.
Loi et al (2002) have imaged high-grade glioma using CT, MR and SPECT.
They showed good correlation between the SPECT and MR functional images and
an advantage over the CT determination. Possibly even the number of clonogenic
cells could be determined to guide IMRT planning.
6.2.4.5 Increased protection of structures
Liu et al (2000) have incorporated functional MRI into radiation treatment
planning. For three patients, two MR datasets were collected. The first was
a fairly standard gadolinium-enhanced set of images and the second was a
functional MR image corresponding to situations in which the patients perform
some task to stimulate the part of the brain required to pursue that task. They
describe these as motor paradigms and visual paradigms (figure 6.5). The
functional MR data were then registered using Interactive Data Language (IDL)
with the gadolinium-enhanced MR data and in turn these were co-registered
with x-ray CT data using the Xknife treatment-planning system. Plans were
then prepared both using the functional MRI data and ignoring it. The dose
distributions to the target and to the so-called eloquent cortex were compared.
It was found that, taking an average over the three patients, the dose to the
eloquent cortex decreased by 32% from 2.4 to 1.63 Gy whilst the dose to the
target remained within acceptable boundaries at a mean of 21.13 Gy.
Liu et al (2000) reviewed other work in which functional MRI has been
shown to give a different image of the tumour. They pointed out that these
previous papers did not investigate the radiation therapy consequences. This is an
important development of the use of multi-modality imaging for radiation therapy
treatment planning. The application was to stereotactic radiosurgery not IMRT.
Garcia-Alvarez et al (2003a, b) have conducted functional MRI to establish
the location of the eloquent cortex which was then protected during IMRT
planning. The scan was made as the patient performed right-handed fingerthumb
241
Figure 6.5. The fMRI brain map (left) and dose distribution (right) for a patient conducting
a finger-tapping motor paradigm. The activated area is at the motor cortex (pointed by
the arrow), which is anterior to the tumour. Right: The prescribed dose distribution from
fMRI-aided treatment plan is plotted on and around the tumour. (From Liu et al 2000.)
opposition. Beavis (2003a, 2004a) and Beavis et al (2003) have also reported on
guiding conformal avoidance via functional imaging.
Borrosch et al (2000) have incorporated functional MRI in the treatment
planning of CFRT for re-irradiation of malignant brain tumours. The aim of the
study was to make use of the MR data to provide information on the function
of the brain that it was required to preserve. For visual stimulation, opening
and closing of the eyes was used as an activation and control condition and
for language stimulation silent word generation was alternated with rest. The
functional MRI data were then fused with x-ray CT data for treatment planning
to irradiate tumours avoiding these functional regions of the brain.
6.2.4.6 Monitoring the response to radiotherapy via MRI
Pickett et al (2000a) have shown that MR spectroscopic imaging allows
differentiation of normal and malignant prostate tissue based on the cholineplus-creatine to citrate ratio. This provides a means of monitoring tumour
response for between one and four years after IMRT for prostate cancer. The
treatment technique used raised the dose of the dominant intraprostatic lesion
above that of the rest of the prostate and 46 patients were treated with forwardplanned segmental multileaf collimation and also re-planned with inverse-planned
242
segmental multileaf collimation. The former led to a more homogeneous dose

distribution and the latter led to a dose distribution which was conformed better
to the prostate and led to high doses within the dominant intraprostatic lesion of
130%. High-dose regions within the prostate appear to be associated with a more
dramatic resolution of metabolic activity as demonstrated using MR spectroscopy.
Pickett et al (2000b) have made use of gold seeds placed in the prostate and
ultrasound guidance and an endorectal balloon to align MR images with CT
images. Using IMRT of the prostate and seven fields, it was shown that the dose
to OARs could be reduced.
6.2.5 Use of functional information from SPECT and PET for treatment
planning
6.2.5.1 Generalities and prostate imaging
The accurate delineation of the target is crucial for the performance of CFRT.
Van Herk (2000) has shown that IMRT requires multimodality image fusion
for treatment planning, allowing the integration of MRI, PET and SPECT into
the treatment-planning process. PET has proven itself as a sensitive indicator
for lung tumours and lymph nodes and it is also useful to differentiate tumour
from atelactasis. PET can also measure lung damage. Moya Garcia (2001)
has shown how the use of PET can yield PTVs which are more appropriate
for the performance of IMRT. In this section are reviewed reports of the use of
SPECT and PET to assist target definition. Ciernik et al (2002) have shown that,
when CT and PET data are fused in the planning of IMRT, the GTV delineation
changed in over 53% of the cases. The changes depended on the tumour site
and were not always in the same direction. The importance of PET in aiding
the planning of IMRT has also been emphasized by Chapman (2003), Kessler
et al (2003) and Cosgrove et al (2003). Van der Heide et al (2003) have used
PET imaging in oncology to upstage or downstage disease prior to treatment
planning. Treatment planning was based on geometrical registration of CT and
MRI data and subsequently the matched PET data were added in order to review
the contour delineation. Miften et al (2004a, b) and Das et al (2004) have assumed
that the active clonogenic cell density is proportional to 18 fluorodeoxyglucose
(FDG)-PET-generated functional images. This information has been used to
create functional-dose-volume histograms (see Lu et al 1997) for the lung and
functional EUD (fEUD). It was proposed that this gives a more realistic basis for
plan comparison. The technique could equally be applied for the prostate.
Xing et al (2002a) have shown how to obtain inhomogeneous dose
distributions within the PTV when the abnormality level (AL) of disease in
the PTV can be established through functional medical imaging (figure 6.6).
The dose to the PTV is guaranteed a particular minimum value and then
the use of an assumed linear relationship between dose and AL determines
the whole needed dose distribution. A similar assumed relationship can be
243
Figure 6.6. IMRT treatment plan for a malignant glioma case. Three action levels are
shown in (A). The isodose distribution is shown in (B). The sensitive structure and target
DVH for different metabolic action levels are shown in (C). (From Xing et al 2002a.)
used to differentially spare normal tissues. The scheme operates on a voxelspecific basis. The main difficulty with this approach is that the required
biological information is generally not known. The authors have proceeded with
assumptions; but this establishes a framework for the future incorporation of
functional imaging. For example, Xing et al (2002b) have shown that when PET
and MR spectroscopic imaging were incorporated into the planning of IMRT, it
was technically feasible to produce deliberately non-uniform doses according to
the functional requirements. Xing et al (2004) have made functional images of the
prostate using MR spectroscopy and then registered these data to CT using thinplate splines as proposed by Bookstein (1989). This allows the determination of
the abnormality level. Miften et al (2004a,b) used such functional abnormality
data to create functional DVHs.
The introduction of IMRT requires a re-evaluation of the complete chain
of radiotherapy processes. The intention at the William Beaumont Hospital,
Royal Oak, was for more than 50% of all patients to be treated with IMRT
from 2001. The goals of IMRT are to dose escalate, to reduce toxicity and to
improve the quality of life. IMRT should be linked to adaptive radiotherapy,
accelerated radiotherapy, image guidance and active breathing control. The
way they are doing this is to identify a lead physician and physicist and the
resources in personnel needed for each target for IMRT. For prostate (see also
Wavelength 2001c), adaptive radiotherapy and inverse planning will be based
on HYPERION and PINNACLE. For lung, active breathing control will be
combined with inverse planning and for breast, techniques are being developed
for whole breast irradiation, regional node irradiation and quadrant irradiation.
At Tubingen also, the link between IMRT and image-guided radiotherapy has
been emphasized. Wong (2000) is using PET to image the washout of the 15 O
positron-emitting isotope when made by in-situ activation with 29 MV external
photon irradiation. This washout serves as a surrogate of perfusion. Analysis
244
of time-binned images clearly showed regions of rapid and slow washout in the
tumour of a canine patient. It was emphasized that the study of tumour biology
with PET imaging is very much in its infancy.
Ling (2001c) has emphasized the role of the Memorial Sloan Kettering
(MSK) Cancer Center in New York in having treated over 1100 patients with
IMRT since October 1995. A decrease of functional tumour volume measured
with FDG-PET as the dose built up throughout the course of treatment was shown.
For the future, it was suggested that imaging biological information can yield
maps of tumour hypoxia, predictive radiosensitivity and genotypic immunology
to feed a better tumour definition for CFRT and IMRT. Ling (2001c) offered the
soundbite: 2001 A Space Odyssey will become 2020 Vision. Both MSK
and the Fox Chase Cancer Center (FCCC) observed a dose dependence of rectal
grade-2 or -3 toxicity and, as the dose escalated (Price et al 2003), the technique
was changed to keep this within acceptable values. Ling (2003) has begun to
investigate the use of functional images for identifying areas of hypoxic tumours.
It was shown that the ratio of choline-plus-creatine to citrate is a good indicator
of Gleason score and that regions in which the choline-to-citrate ratio is greater
than two indicates hypoxic areas that would benefit from increased dose in an
inhomogenous dose distribution, for example for dominant intraprostatic lesions.
Hypoxic cell markers which also relate to changes in TCP, for example the
markers 18 F fluoromisonidazol (FMISO), 124 I iodo azomycin galactoside (IAZG)
and 60 Cu diacetyl-bis-methylthiosemicarbazone (ATSM) were investigated. It
is important to validate that regions which showed changed uptake of these
hypoxic cell markers are indeed hypoxic and Ling (2003) is using measurements
of oxygen in mouse tumours to correlate with the distribution of uptake of these
hypoxic cell markers. Somewhat tongue-in-cheek, Ling (2003) has pointed out
that the application of imaging to planning radiotherapy is having something of
a renaissance since Rontgen, Becquerel and Curie all got together (figuratively)
in the late 1890s to apply the use of x-rays to planning radiotherapy. Lewellen
(2001) has also emphasized that molecular imaging using PET can define volumes
for boost protocols in radiation therapy.
Dehdashti et al (2003) have shown that there is increased uptake of 60 Cu
ATSM in hypoxic tumours. They found that this was a good indicator of
disease-free survival, this being lower for hypoxic tumours due to their increased
radioresistance. PET images of 60 Cu ATSM could also provide maps of hypoxia
allowing the generation of deliberately inhomogenous dose distributions via
IMRT.
6.2.5.2 Head-and-neck imaging
Nishioka et al (2000) have evaluated the impact of 18 FDG-PET imaging
combined with x-ray CT and MRI for radiotherapy planning of head-and-neck
tumours. They found that, among primary tumours, the volume of the primary
tumour was changed in two out of 16 patients by making use of the PET data,
245
Figure 6.7. An example of 18 FDG-PET/MRI fusion for nasopharyngeal carcinoma.

An automatic multi-modality image registration algorithm enables accurate registration
between the two modalities using the brain as an internal reference. Anatomic locations
of 18 FDG-PET uptakes which were unclear on the original PET images (left), became
clear on the fusion images (right). (Reprinted from Nishioka et al (2002) with copyright
50% larger in a case of carcinoma of the base of the tongue and 35% smaller in a
case with tongue cancer. Treatment planning was improved for 10 out of the 16
patients by the introduction of the PET data coregistered with CT and MR data.
Nishioka et al (2002) have fused 18 FDG-PET and MRI/CT image data for
radiotherapy planning for oropharyngeal and nasopharyngeal carcinomas. They
studied 12 patients in the former category and nine in the latter. Except for three
cases with superficial tumours, all primary tumours were detected by PET. The
GTV volumes for primary tumours were not changed by image fusion for 19
cases, were increased for 1 case by 49% and decreased for another case by 45%.
Parotid sparing became possible in 15 patients whose upper neck areas near the
parotid glands were tumour free by PET (figure 6.7).
Daisne et al (2002) have shown that CT, MR and FDG-PET lead to
quite different definitions of GTV in head-and-neck squamous cell carcinoma.
Klabbers et al (2002) have registered PET and CT data for head-and-neck tumours
in which the patient was immobilized inside a mask to provide a method of
image registration. Paulsen et al (2002) have shown that areas of hypoxia in
head-and-neck tumours can be identified using the hypoxic marker FMISO. By
combining PET and CT data, volumes in which the dose need to be boosted due
to hypoxia were identified. Nusslin et al (2003) have used HYPERION to plan for
246
inhomogeneous dose painting using the information from fluorine-misonidazole

PET imaging of a head-and-neck tumour.
6.2.5.3 Lung imaging
Ung et al (2000) have shown that FDG-PET significantly alters treatment
planning in cases where atelectasis or unsuspected node involvement exists. The
combination of PET and x-ray CT data was useful for accurate determination
of the GTV and minimizing geographical miss and was a fundamental step in
treatment planning for lung cancer. Benard et al (2000) have shown that FDGPET imaging for suspected lung cancer can improve the treatment of radiation
therapy for the lung. Erdi et al (2000) have also shown that the use of FDGPET in conjunction with x-ray CT improves the treatment planning of non-smallcell lung cancer. Transmission PET images were acquired simultaneously with
emission PET images. The PET transmission images were registered with the xray CT images in the treatment-planning system using either manual or automatic
methods which in turn leads to the consequent registration between x-ray CT
and the PET emission images. Six patients were evaluated and, for three of the
patients, the PTV was increased about 10% to include nodal disease. In one
patient, PET detected a second lesion in the lung. The incorporation of PET data
improved the definition of the primary tumour and should therefore improve the
chances of achieving local control. Balogh et al (2000) also showed that coregistered FDG-PET images with x-ray CT improved the tumour localization for
non-small-cell carcinoma of the lung. Schmuecking et al (2000) have also shown
that whole-body PET imaging can improve the target definition for non-small-cell
lung cancer. Findings for 31 patients were recorded. It was found that the use of
PET reduced the PTV by between 321% compared with conventional x-ray CT
imaging, largely due to the exclusion of atelectasis resulting in a smaller PTV.
However, in some patients, the PTV increased because lymph node metastases
which were inconspicuous in CT were also determined using the PET imaging.
Bradley et al (2004) found the use of PET-CT changed the treatment plan in
over 50% of cases. Christian et al (2003b) and Partridge et al (2003) have
correlated CT scans with SPECT perfusion scans for patients receiving treatment
for lung cancer (figure 6.8). They were able to show that beam orientations could
be selected which reduced the irradiation of healthy (i.e. perfused) lung tissue.
Orientations could be chosen which were passing through already damaged lung.
Das et al (2003, 2004) have used FDG-PET scans to characterize and define a
lung tumour and SPECT to characterize lung function simultaneously. They were
thus able to define beams which spared normal lung whilst treating the tumour.
Deliberately inhomogeneous PTV dose distributions were created assuming that
the PET uptake correlated to tumour proliferation. Simultaneously the SPECT
uptake guided avoidance of normal lung with good perfusion. Paulino and
Johnstone (2004) expressed concerns about FDG-PET imaging of lung tumours:
(i) what threshold should be used for identifying a tumour?
247
Target
Figure 6.8. A fused image of 3D SPECT perfusion of the lung and 3D CT showing the
areas of non perfusion in the upper pole of the left lung. Beams directed at the tumour via
this region preferentially may be better than others at sparing normal lung. (Courtesy of
Dr M Partridge.) (See website for colour version.)
(ii) What is the effect of poorer spatial resolution than in x-ray CT?
(iii) Who will contour the volumes and with what prior experience?
(iv) will patient respiration not be a bigger problem to solve than concern over
tumour definition?
6.2.5.4 Para-aortic lymph node (PALN) imaging
Mutic et al (2003) and Esthappan et al (2004) have used PET to create a
GTV for involved para-aortic lymph nodes (PALNs) for four cervical cancer
patients (figure 6.9). FDG is taken up by cancer cells because of the high
glucose use of tumours arising from increased blood-flow, increased glucose
phosphorylation and decreased dephosphorylation and increased uptake of cell
membrane transporters. IMRT for PALNs has been designed to escalate the dose
beyond that to the PALN bed defined as CTV. Beitler (2003) has emphasized
the importance of having PET images to guide planning of advanced cervical
cancer and Miller and Grigsby (2003) indicate its relevance also to brachytherapy
planning.
248
Figure 6.9. Axial PET and CT images showing a positive paraaortic lymph node (PALN)
and kidneys. The left-hand image was obtained by PET imaging of fluorodyoxyglucose
using a ECAT EXACT scanner from Siemens. The PALN is outlined as well as the uptake
in the kidneys. The right-hand image shows the corresponding CT scan with the positive
PALN superposed. (Reprinted from Mutic et al (2003) with copyright permission from
Elsevier.) (See website for colour version.)
6.2.5.5 Combined PET-CT scanning

Baumert et al (2002) have used a PET-CT combined scanner to image tumours
of the head-and-neck, lung, gynaecological or recto-anal carcinoma. They found
that, in six out of the 30 patients, the use of the PET data led to significant changes
in the dose to the tumour, dose to the lymph nodes and in the extension of the GTV
whereas there were no significant differences in the PTV.
6.2.5.6 Imaging to overcome breathing-motion effects
Mah et al (2000) have investigated the hypothesis that spiral CT may not
accurately represent the time-averaged position of thoracic tumours during
radiotherapy. Spiral CT can be acquired in about 0.7 s whereas the breathing
cycle is about 4 s. In addition, PET with FDG can acquire data over 45 min. They
concluded that PET and non-spiral CT may better represent the time-averaged
position of the tumour. This study was described in detail by Caldwell et al
(2003). Measurements of target volume were made by spiral CT and by PET of
a phantom with three fillable spheres of known volume. These were measured on
an oscillating table. Using CT data, a margin of 15 mm had to be added to ensure
no beam geographical miss whereas the volume determined from (lengthy) PET
led to good geographical hit and lower dose to normal structures. Van Sornsen de
Koste et al (2003a) also showed that the use of a single slow CT scan gave a more
faithful PTV than would be obtained from a rapid spiral CT scan. Averaging over
six rapid spiral CT scans was an alternative way to get the optimal PTV but much
more time consuming. They used the slow-scan CT to create a PTV averaged
249
Figure 6.10. The upper panel shows co-registered rapid (left) and slow (right) CT scans
of a tumour in the lung. Blurring of the edges of the mobile structure such as vessels
in the heart are obvious on the slow scan. The middle panel shows CT scans showing
the corresponding contoured GTVs. The lower panel shows the projection of all contoured
GTVs on each type of scan. The conclusion is that the slow-scan CT can create an averaged
PTV over the breathing cycle more conveniently than averaging over six rapid spiral CT
scans. (Reprinted from van Sornsen de Koste et al (2003a) with copyright permission from
Elsevier.)
over the breathing cycle for lung tumours and a single fast spiral CT scan for
visualizing other mediastinal structures (see also section 6.10.12.2) (figure 6.10).
Low et al (2001b) have shown that thoracic movement can significantly
blur the distribution of activity in a lung tumour as viewed by a PET scanner.
They proposed to use active breathing control (ABC) to minimize the motion
250
contribution and to subsequently allow the patients to be treated with IMRT for
upper thoracic tumours.
Humm et al (2003) have pointed out that the different data acquisition times
for PET and CT can lead to concerns for registration. For this reason, gated PET
or respiratory synchronized PET may be the solution.
6.2.6 Use of pathology specimens to compare with GTV and PTV

Teh et al (2003) have compared the pathologic prostate volume (PPV) determined
from postoperative prostate pathology to the CT-based GTV and PTV for 10
patients selected from a study of 712 patients who underwent prostatectomy. They
found that the GTV was on average twice as large as the PPV. The average PTV
was 4.1 times larger than the PPV when a 5 mm margin was used. They thus
concluded that the extent of prostate disease including extracapsular extension
was well covered during CFRT. The rest of this study is a detailed analysis of the
specifics of extracapsular extension.
Gregoire et al (2003) have also carefully correlated the GTV determined
from CT, PET and MR for squamous cell carcinoma of the head-and-neck with
pathology excision specimens. They found the very interesting (and worrying?)
conclusion that the 3D imaging modalities overestimated the GTV. The PETdefined GTV was closest to pathology reality. Also all imaging modalities missed
small microscopic extension.
Nabid et al (2002) correlated the size of tumours reported by total-body PET
scans and pathology reports for resected tumours but no detailed results were
reported.
6.3 New inverse-planning methods for IMRT

Methods for inverse planning for IMRT have been developed now for about
15 continuous years (see reviews in Webb 1993, 1997, 2000d). At least for
the first half of this period, the techniques were almost exclusively those based
on minimizing a simple dose-based cost function by the analytic technique,
gradient descent or the iterative technique, simulated annealing. Since about
1997, dose-volume constraints have been incorporated into cost functions. These
are generally of the type let no more than x% of the volume of an OAR receive
more than y% of the target dose. Lian et al (2002) have introduced a probabilistic
dose prescription as the basis for IMRT planning. The idea is to specify the
preference for a particular dose that a voxel has in terms of a probability function
rather than a strict single value. By replacing the rigid dose prescription by a
probability function, it was shown that plans can be improved. Bedford (2003) has
reviewed optimization and inverse planning for IMRT describing the fundamental
concepts and practical factors in its implementation.
251
6.3.1 Gradient-descent inverse planning

Wu and Mohan (2000) have developed an algorithm for inverse planning which
has been made to operate inside the ADAC treatment-planning system. It
uses dose-based and dose-volume-based cost functions and a gradient-descent
technique. The code allows overlap regions between PTV and OAR as PTV, OAR
to be specified or as a separate region. They showed the effect this can have on
tailoring the dose distribution although only three clinical examples were shown.
6.3.2 Simulated annealing inverse planning
Sham et al (2001a, b) have reformulated inverse planning using volume
constraints instead of dose-volume constraints. Instead of the gradient method,
the simulated annealing method to minimize a volume-based cost function was
also used. They find this advantageous for prostate cases and show that it
overcomes difficulties associated with the dose-volume constraint.
6.3.3 Equivalent-uniform-dose-based inverse planning
Wu et al (2000a, 2002b) have used the equivalent-uniform-dose (EUD) concept as
the basis of a cost function in inverse planning and compared the results with those
generated using dose-based and dose-volume-based criteria (figure 6.11). It was
argued that dose-volume constraints are simplified surrogates of the underlying
biological effects. Mathematically, dose-response functions are highly degenerate
functions of dose-volume combinations and therefore of dose distributions. It was
argued that the use of the EUD, whilst remaining in the dose domain, makes it
easier for the clinician to specify the requirements for plan evaluation. The EUD
mimics the reality of dose response more closely than the simple dose-volume
parameters. A case of prostate IMRT and another case of head-and-neck IMRT
were examined and it was noted that improvement resulted from the use of EUD.
It was found that the EUD-based optimization improved the sparing of critical
structures but decreased PTV dose homogeneity. The target was included as a
normal structure in order to circumvent this and to constrain the maximum dose
to the target.
Wu et al (2003b) have extended the IMRT optimization technique based
upon the generalized equivalent uniform dose (gEUD) concept. The main
advantage proposed is that the gEUD technique always leads to a plan which
improves on those based on dose-volume constraints. The extension is a twostage process; if the gEUD-based optimization does not totally meet all the criteria
then a subsequent second stage of dose-volume-based optimization takes place.
The study considered two planning situations and showed improved dose-volume
histograms based on the new hybrid approach.
Bortfeld et al (2002c) have also developed a new inverse-planning technique
based on evaluating the EUD in an organ and looking at the trade-offs between
different entities. Thieke et al (2003ac) have used EUD as the base of the cost
252
Figure 6.11. Dose distributions in a transverse plane through a head-and-neck tumour

for (a) dose-volume-based objectives, (b) EUD-based objectives with dose homogeneity
constraint, (c) EUD-based objectives without dose homogeneity constraint. (From Wu et
al 2002b.)
function of a projection-onto-convex-sets iterative inverse-planning technique

inside the KONRAD system. Schwarz et al (2003) have made a series of
prostate IMRT plans based on EUD and using different values of the EUD-volume
parameter n showing that plans vary according to this value.
Lian and Xing (2004) have further developed IMRT optimization based on
the EUD model. They have, instead of using a fixed value of the a parameter in
the formula for the EUD, included the possibility to vary this parameter according
to a probability distribution (in their study a Gaussian distribution). The objective
function was then reformulated for each structure in the presence of an uncertainty
in this a parameter in terms of joint probabilities and an overall objective
function was formulated that takes account of the uncertainty in the parameters.
Using two model problems, it was shown that the outcomes are very dependant
on the form of the distribution of this EUD parameter a. For two planning cases,
four extreme forms of distribution were considered.
Schreibmann et al (2004b) have developed beam-orientation optimization
for IMRT based on EUD constraints.
6.3.4 Maximum entropy inverse planning

Wu and Zhu (2001b) have introduced a maximum entropy (ME) method for
the planning of CFRT. The philosophy behind the technique is that maximizing
entropy minimizes information in the treatment plan and this corresponds to the
maximum homogeneity of dose and the closest approximation of the calculated
dose to the prescription dose. The technique relies on the fact that fluence
distributions are positive and additive. The Gull and Skilling (1984) definition

of entropy was used and is given by

NT

Dk (X)
PTV
Dk (X) Dk Dk (X) log
E(X) =
DkPTV
k=1
253
(6.3)
where DkPTV is the expected dose delivered at point k in the PTV, NT is the number
of points in the PTV, Dk (X) is the calculated dose distribution at the same point
k.
A cost function was constructed which includes a weighted negative leastsquares error term for dose constraints on OARs in addition to the entropy term.
This second term is given by
U (X) =
Ns
(Dk (X) DkOAR )2
(6.4)
k=1
where DkOAR is the maximum allowable dose at point k in the OARs and Ns is the
number of constraint points in the OARs. The cost function becomes
Q(X) = E(X) U (X).
(6.5)
A NewtonRaphson algorithm was used to solve the minimization of the

constructed cost function. If the quantity in the cost function is high, then
protection of the OARs is relatively more important than the elimination of the
tumour cells in the PTV. Conversely, if the quantity is low, the homogeneity
of dose distribution in the PTV is much more important than that of the OARs
and the entropy in the PTV dominates the cost function. The technique was
implemented inside the PLUNC treatment-planning system at the University of
North Carolina, Chapel Hill, and its effectiveness was demonstrated compared to
manual planning for CFRT of two difficult planning cases.
Arellano et al (2000) have developed a variant of the dynamically penalised
likelihood inverse-planning algorithm based on the maximum likelihood
estimator. This accepts few and only clinically relevant pre-optimization
parameters and calculates several alternative optimized solutions for the same
clinical case.
6.3.5 Genetic algorithms
One of the main problems in inverse planning is the selection of the importance
factors (IFs) which govern the importance of dose conformation and/or dose
avoidance and which in turn determine to a large extent the planning outcome.
Wu and Zhu (2001a) and Wu et al (2000b) have developed a technique to
automatically determine both the beamweights (in conventional planning) and the
IFs using a genetic algorithm. The chromosomes at the centre of the algorithm
have as their components the set of beamweights for the set of fields and the set of
254
IFs for the PTV and for the OARs. Then the algorithm progresses by first fixing
the IFs and iterating genetically for the beamweights followed by a stage with
the weights held constant and the iteration being made for the IFs. The process
cycles until the cost function based on these IFs and the prescribed dose-volume
constraints is satisfied or some termination condition is reached. They showed
that this gives better plans than manual plans for three cases. i.e. for satisfied
PTV constraints, the doses to OARs are lower.
Li et al (2003ac) have developed a new form of DAO. The method
simultaneously determines the number of segments per beam, the segment shapes
and their weights. An interim stage involved determining the number of segments
per beam based on the IMB complexity determined by a conjugate gradient
method and then weights by a conjugated gradient method. The shapes for this set
were then optimized using a genetic algorithm in which mutations and crossovers
of selected parts of field shapes were interchanged. The goal was to decrease the
total number of segments needed for a conformal dose distribution; the result is
superior to the use of equally distributing the segments between beams. Roughly
30 segments were determined to be adequate for quite complex planning cases.
Li et al (2004b) have used genetic algorithms to optimize beam directions.
Cotrutz and Xing (2003a,c) have used a genetic algorithm to directly
select and optimize MLC shapes for IMRT (figures 6.12 and 6.13). The three
chromosomes representing an aperture coded for the positions of the left and
right leaves respectively and the weight of radiation for segments so defined.
This algorithm led to highly conformal dose distributions with fewer segments
than would have been obtained by segmenting beams obtained from a bixel-based
optimization scheme. No topological restrictions were applied other than these
imposed by the MLC itself. It was observed, as expected, that conformality
increased with increasing the number of prespecified segments, convergence
being correspondingly slower. For a prostate case, typically five to seven apertures
proved sufficient for acceptable conformality.
6.3.6 Single-step inverse planning
Chuang et al (2003) have presented a new technique for inverse planning which
determines the beam intensity profile in a single step instead of the more common
deterministic or stochastic iterative algorithms. The single-step method utilizes a
figure of merit. The figure of merit (figure 6.14) is effectively the ratio of the
dose deposited in the tumour to the dose deposited in the normal tissues for each
ray and the first step in the process is to create the beam profiles with the bixel
values simply set to the figure of merit for each ray. This figure of merit contains
a number of factors which allow it to work more effectively. For example, there
is a dose offset term which enables the dynamic range of the figure of merit to
be limited. If this dose offset is set very high, the beams would automatically
become almost uniform. There is a term representing the intersection of each ray
with each pixel and other terms control weighting factors applied to the tissue,
30%
255
30%
60%
80%
60%
PTV
95%
95%
80%
OAR
30%
60%
95%
30%
80%
60%
30%
Figure 6.12. Isodose distribution for a C-shaped tumour optimized using seven apertures
per beam and nine equispaced 6 MV photon beams and a genetic algorithm for creating
segments. (From Cotrutz and Xing 2003c.)
penalty functions for dose to normal tissues and a so called star-pattern correction
matrix that corrects for the backprojection star which is created when beams
from different directions intersect and which naturally leads to a higher density of
intersection close to the isocentre than further away.
The second stage of the algorithm is to maintain the form of the fluence
modulation profile and then to weight each whole profile accordingly in order to
maximize the uniformity of dose to the tumour. This is an iterative process but is
much quicker than changing the individual beamweights of bixels. It was shown
that, compared to a commercial algorithm (CADPLAN), the method yielded a
lower rectal dose but a greater inhomogeneity in the prostate dose distribution.
The smoothness of the intensity profiles was controlled by dose offset in the
calculation of figure of merit. The authors proposed an extension to the concept
whereby dose-volume constraints and other hard dose constraints will be applied
to individual beam elements. The basic idea is that if each individual beam fulfils
the constraints, then the linear combination of multiple beams will also satisfy the
constraints. This technique appears to have much merit.
6.3.7 Simulated particle dynamics
Hou and Wang (2001) and Hou et al (2003a) have developed an optimization
algorithm for IMRT based on simulated particle dynamics. The algorithm is
256
Figure 6.13. Axial slice of the final dose distribution optimized with seven equispaced
6 MV photon beams, each with three segments created using a genetic algorithm. (From
Cotrutz and Xing 2003c.)
based on an analogy between the procedure of searching for the optimum intensity
profiles in IMRT and the relaxation of a dynamic system of many interacting
particles. Hard constraints that require non-negative beamlet intensities can be
implemented and also constraints that the dose at any voxel in an OAR cannot
exceed a maximum tolerance are implemented as semi-transmittable potential
barriers of infinite height. The process takes place without the need to specify
IFs.
Hou et al (2003) used a genetic algorithm to select beam orientations
and a simulated dynamics method for optimization of beam fluence at each
chosen set. They confirmed earlier results from Stein et al (1997) that some
IMBs preferentially traverse OARs. Also fewer beams were better than a large
number of beams as candidates for optimization of beam direction. The preferred
orientations also depended on the prescription goals. It was also found that, for a
complicated clinical case, IMRT with five or seven well-optimized directions gave
better results than nine equally spaced beams, a result known from Rowbottom et
al (2001).
257
Figure 6.14. An example of a plot of figure of merit for four orthogonal angles of beams
irradiating the prostate. The dose offset was set at 100 units and the rectum was weighted
by a factor of three with respect to target. (From Chuang et al 2003.)
6.3.8 Optimization of surrogate parameters in beam space

Markman et al (2002) have presented a very interesting and quite novel technique
for inverse planning. Usually conventional IMRT is separated into two distinct
problems, that of inverse planning and that of interpreting the inverse-planninggenerated IMBs into practical IMRT delivery techniques. The essence of this
study was to couple together the problems by ignoring direct bixel optimization
and instead optimizing a set of parameters which relate to the bixels but into which
can be built delivery constraints.
The first of the techniques investigated was to optimize the fluence at a set
of anchor points, these being more widely spaced than the conventional bixels.
The number of anchor points is thus fewer than the total number of bixels and the
bixel intensities are interpolated using functions into which physical properties
can be factored. The types of interpolation used were:
(i) nearest-neighbour interpolation (which is effectively the same as optimizing
over the bixel elements directly at a lower bixel resolution);
(ii) linear interpolation which ensures bounded gradients and which should be
less sensitive to positioning errors and, finally,
(iii) cubic interpolation which may result in steeper gradients but reduce higherorder derivatives.
258
Figure 6.15. An example of determining the fluence profile (1D) for IMRT using radial
basis functions. The number of exponential distributions, their centres and widths are
pre-defined by the user. The optimization algorithm then determines the height of each
function to calculate the fluence profile. The bixel intensities are then sampled from the
resulting function shown as a dotted line in the figure. (From Markman et al 2002.)
The fluence at the anchor points was directly determined by the optimization
method and the remaining bixel fluences were then calculated through
interpolation.
A second form of indirect optimization of a parameter set is to expand an
IMB as a set of positive radial basis functions (RBFs). The RBFs used were
exponentials centred around different positions and of different widths and in the
study ten such RBFs were used spaced uniformly in the open field. By optimizing
the amplitude of these RBFs, only 10 optimization points per beam entered into
the calculation. The final IMB will consequently have been constructed on a
much finer grid (figure 6.15). This scheme optimizes a number of parameters
which are less than or equal to the number of fluence bixels. It also enables
favourable properties of the beam to be incorporated (e.g. direct smoothing) and,
in principle, can speed up the optimization by a factor of between two and three.
The optimization itself takes place in a fairly traditional way using a quadratic
dose-based least-squares objective function with relative importance weights.
Two clinical cases were investigated with five equally spaced 6 MV photon
beams for each. Conversely, standard optimization was also performed and
comparisons were made between the two newly introduced techniques and the
standard method. Results showed typically a factor of two speed-up in run time
and the reduced parameter cases preserved the general shape and magnitude of
the dose distributions whilst introducing the required smoothing into the beam
profiles. It was also shown that, by deliberately misaligning the beam profiles by
a millimetre, that the smoother profiles were less sensitive to misalignment.
259
6.3.9 Comparison of inverse-planning techniques

The effects of three (Dejean et al 2000b) and four (Dejean et al 2001a)
regularization methods for performing IMRT using the singular value
decomposition (SVD) technique have been shown. Dejean et al (2000c) have
compared analytic SVD with the iterative gradient technique. For the gradient
technique, the parameter selection is very user dependent whereas SVD is a more
natural inverse process. Dejean et al (2001b, c) have investigated the use of
biological cost functions to compare with the least-squares cost function used
in their SVD inverse-planning algorithm. Dejean et al (2000a) have used receiver
operating characteristic (ROC) analysis to evaluate IMRT. Rousseau et al (2002)
identified that, of the two inverse-planning techniques embodied in the Dosigray
inverse-planning module, the SVD technique gave a better homogeneity in the
PTV but worse OAR results when compared with an iterative projected-gradient
technique. The application was to a nasopharynx tumour attempting to spare brain
stem and cord.
Wu et al (2001a) have compared a fast simulated annealing stochastic
planning algorithm with a deterministic inverse-planning algorithm and showed
that they both generate similar optimal solutions for the cases studied. They
studied the effectiveness in optimization, the convergent speed and the sensitivity
to the initial guess. It was claimed that this was the first time such a comparison
had been explicitly made.
Llacer et al (2001) have carried out a very thorough comparison of five
inverse-planning algorithms. These were the dynamically penalized likelihood
(DPL) algorithm, a variation on this, a new fast adaptive simulated annealing
algorithm, a conjugate gradient method (of Spirou and Chui 1998) and a Newton
gradient method (of Bortfeld et al 1990). These were applied to test phantoms
and patient cases. All five can yield similar results but they conclude that the
DPL algorithm has advantages (figures 6.16 and 6.17). Filtering of IMBs was
specifically discussed as well as the need to avoid hot beamlets.
Crooks and Xing (2001) have developed inverse planning techniques which
use linear algebra. They compared the outcomes with the results from the
commercial CORVUS system for several clinical cases. They concluded that
very similar results were obtained with the new technique but with much less
computer time and thus advantage. Linear programming was also developed by
Romeijn et al (2003). Luo et al (2002) have also developed a linear algebraic
approach that, instead of minimizing the quadratic difference between the desired
and computed doses, minimizes an objective function which is the quadratic
summation of negative beamweights. Goldman et al (2004) have described an
inverse-planning algorithm called FIDO (fast inverse dose optimization), which
works out the beamweights for IMRT by solving linear equations in which the
conventional quadratic cost functions have added to them a term which forces the
beamlets to stay positive. The linear equations so generated are then solved by
fast matrix inversion methods. A patent is pending on this planning method.
260
Figure 6.16. Plane containing the isocentre (cross) of a cavernous sinus meningioma case
showing two of the specified OARs (left optic nerve and brain stem) and the PTV. (From
Llacer et al 2001.)
Figure 6.17. Dose distribution resulting from optimization of the cavernous sinus
meningioma case by the dynamically penalized likelihood algorithm. Here we see just
a part of the plane shown in figure 6.16 in order to focus attention on the PTV and
the high-gradient borders with the OARs. Seven isodose levels are shown in patches of
different grey from 40% to 100% dose. (From Llacer et al 2001.)
261
Morrill et al (2001) compared, as black boxes, two inverse-planning systems

for IMRT. They found the two (unnamed) systems comparable when static stepand-shoot fields were delivered with 1 cm-wide leaves.
Richter et al (2000) have compared a rotation technique with IMRT and
found the latter superior for concave targets. Calculations were performed using
the HELAX treatment-planning system.
Yi et al (2002) have developed a radiotherapy research tool box for IMRT
optimization in which they compared, and showed to be similar, the outcomes of
simulated annealing and the gradient-search optimization technique.
6.3.10 Features and comparison of commercial planning algorithms
Cozzi et al (2001b) have compared the HELAX treatment-planning system
and the Varian CADPLAN-HELIOS treatment-planning system for IMRT.
They studied advanced head-and-neck cases and found that comparable dose
distributions could be obtained provided the step-and-shoot inverse plans have
between seven and eight intensity levels. With less than this, the dose distributions
calculated were clinically unacceptable. Fogliata et al (2003) extended this study
to include a third system, the CMS FOCUS, again finding all three systems gave
comparable results provided common planning strategies and constraints were
applied.
ODaniel et al (2002) have compared inverse planning using the ADAC
PINNACLE system and using CORVUS. They found that, in general, PINNACLE
IMRT plans produced a more homogenous target coverage with more efficient
treatment delivery than CORVUS but that the CORVUS plans were more
conformal and provided more critical-structure sparing.
Dogan et al (2001b) have compared IMRT plans created with the NOMOS
CORVUS and with the CMS FOCUS treatment-planning systems. Plans created
on both systems used seven non-opposed coplanar beams for the following
sites: brain, lung, head and neck, nasopharynx and prostate. Critical-structure
dose constraints were site-specific and the same for both treatment-planning
systems. CORVUS uses simulated-annealing optimization and finite-size pencilbeam dose-calculation algorithms. CMS FOCUS uses simultaneous iterative
inverse-planning optimization and superposition dose-calculation algorithms.
Both systems produced satisfactory target coverage. CMS FOCUS demonstrated
somewhat more inhomogeneous target dose distributions but better sparing of
dose to critical structures.
Rosello et al (2001) have shown significant discrepancies between the
IMRT predictions of a HELAX treatment-planning system and experimental
measurements when small fields are placed in the corner of large areas.
Schwarz et al (2002) have considered the importance of accurate dose
calculation outside the beam edges in IMRT treatment planning. For 10 patients
(five prostate and five head-and-neck cases) they showed the UMPLAN planning
262
system tends to overestimate the dose whereas PINNACLE tends to underestimate

the dose for prostate plans. This was not considered significant.
Samuelsson and Johansson (2003) have made a systematic study of the
Varian HELIOS planning system for planning head-and-neck treatments. They
found conformality always increased as the number of beams increased, that
varying the collimator angle had very little effect and that to achieve dose-volume
specification, lower dose-volume constraints should be set.
Martin and Hachem (2003) have commented on the lack of superposition
of dose when multiple field segments are added together to create an intensitymodulated pattern. They note that the HELAX planning system effectively adds
together all the components of a field before making a one-shot computation of
dose. However, in practice, the dose is built up from individual segments and,
because of the TG effect and also because of penumbra phenomena, the delivered
dose can differ from the calculated dose in specific regions. This is because
internal penumbra is not taken account of in the HELAX planning system. They
call this a nonlinear accident.
Johansen et al (2003) have compared the TMS and the CORVUS system for
inverse planning noting subtle differences in the generation of minimum dose to
target, being case dependent.
6.3.11 Dependences of IMRT plans on target geometry
Hunt et al (2000) have evaluated the dependence of IMRT on the number of
fields (5, 7 or 19), on the dose constraints set for normal-tissue tolerance and
on the geometrical separation distance between normal tissue and targets. Not
surprisingly, they found that at small target-to-normal-tissue separations, normaltissue sparing is improved as the number of beams increases and target uniformity
is hard to achieve. Their results were used to determine sensible constraints
for inverse treatment planning. Ezzell (2003) has also commented that if the
planner asks for an impossible plan this will at best lead to a suboptimal plan
and at worst may lead to no plan being possible. In short: An inverse plan may
become a perverse plan. He provides a recipe for those who wish to commence
learning about inverse planning. Experience is built starting with simple planning
problems and building to more complex ones.
Hunt et al (2002) have further investigated the effects on target homogeneity
and coverage in inverse planning for IMRT of varying the radius of curvature of
the target, the separation between the target and the normal tissues to be spared
and the tolerances which were set for the normal tissues. This study tried to
achieve a more systematic understanding of how the dose distributions achieved
by inverse planning for IMRT depend on these particular parameters.
The optimizations were designed using a least-squares objective function and
a gradient minimization technique developed at the Memorial Sloan Kettering
Cancer Center and all constraints in this system are so called soft constraints.
To get an understanding of these dependences, a variety of concave targets and
263
normal tissues were simulated, mapping situations commonly encountered in

head-and-neck and other disease sites within a cylindrical phantom.
The results are quite complicated and a good understanding of the behaviour
can only be had by looking at the many figures in the paper (e.g. figure 6.18).
It was found, for example, that relaxing the maximum normal-tissue constraints
led to more uniform target doses and less concave dose distributions. Similarly,
as the PTV-to-normal-tissue separation became smaller, the uniformity index and
the normal tissue dose began to exceed the constraints by increasing amounts.
All of these behaviours are completely understandable as far as the direction of
behaviour is concerned. The strength of the study is that it leads to a strategy
for optimization. In particular, to produce an acceptable plan, the user-specified
values for the target- and normal-tissue constraints had to be more stringent
than the actual clinically imposed dose limits. Secondly, identical constraints
were found to lead to different results depending on patient-specific geometries
of target and normal tissue. The optimization results always exhibited a strong
dependence on the geometric relationships in the patient. It was therefore decided
to operate an iterative inverse-planning technique in which the whole planning
process was put in a do-loop over constraints. Initially the dose constraints
for targets and normal tissues were set to values approximately 10% lower
than the clinical goals and the normal-tissue penalties were equal to or less
than the target penalties, Then after the first optimization and re-normalization,
the criteria were selectively adjusted with progressively smaller modifications
until the changes between successive optimizations were considered clinically
insignificant. The study provided useful quantitative information that guides
treatment planning for IMRT. Kron et al (2004b) similarly showed that the degree
of conformality correlated well with the degree of overlap of PTV and OAR for
helical tomotherapy.
Kapatoes et al (2000) have also pointed to the importance of assessing dose
gradients in IMRT treatment planning. The nature of IMRT is such that it is
characterized by large dose gradients and therefore misalignments of the patient
with the modulated beams can have quite serious consequences in failing to
achieve the planned dose distribution. It was found that gradients above 30%
per cm were associated with interfaces between the target and OARs but that the
gradients were at least half this between target- and normal-tissue boundaries.
Kapatoes et al (2001b) have developed a strategy for correcting the small number
of underdosed and overdosed voxels in a dose distribution that may arise due to
optimization limitations. They call this the fluence adjustment strategy.
6.3.12 Multiple local minima and the global minimum in optimization

Many workers have shown that the use of specific cost functions in inverse
planning leads to the existence of local minima (independently: Deasy, Spirou
and Chui, Llacer, Webb, Rowbottom, Magereras, Mohan and Wu). Stochastic
264
Figure 6.18. Examples of the phantom, target, normal-tissue geometry and beam
arrangements used in the study by Hunt et al (2002): (a) Seven-field beam arrangement
with the 1.5 cm radius of curvature PTV and the 1-cm-radius normal tissue positioned
8 mm from the PTV edge; (b) five-field beam arrangement with the 2.75 cm radius of
curvature PTV and the 2.25-cm-radius normal tissue positioned 8 mm from the PTV edge.
These are two of the software phantoms designed to study the effect of the radius of
curvature and the disposition of the PTV and OARs in the study. (Reprinted from Hunt
et al 2002 with copyright permission from Elsevier.)
search techniques such as simulated annealing have been used to find a global
minimum avoiding local minima.
Wu and Mohan (2001) have, for example, shown that multiple local minima
exist in IMRT inverse plans based on dose-volume constraints and dose-based
objective functions. A particular configuration was studied with 500 different
plans optimized with random initial beamweights and it was shown that the plans
clustered in terms of multiple local minima, some plans being significantly better
than others. It was thus concluded that avoiding local minima in inverse planning
was desirable. However, Wu and Mohan (2002) studied whether in practice the
use of some solution corresponding to a local instead of global minimum would
have any clinical significance. For selected cases they ran a gradient descent
optimization similar to that of Bortfeld et al (1990) hundreds of times from
different starting conditions thus identifying local minima. They analysed cost,
dose-volume histograms, dose-plans and modulation patterns to deduce if there
were any clinically detectable differences. For a phantom case they observed
multiple local minima in cost and beamweight when only two to four beams were
used without intensity modulation. For clinical cases with intensity modulation,
the situation changed and no local minima were observed. They postulated that
265
the existence or otherwise of local minima depends on the geometry, number of

beams and number of bixels and that the more degrees of freedom there were,
the less the likelihood of local minima providing a problem that has clinical
significance. Wu et al (2003a) have discussed the problem of local minima in
IMRT optimization. They have developed a method of intercepts in parameter
space to analyse local minima caused by dose-volume constraints and shown that
many of these local minima correspond to very similar cost functions. However,
occasionally they can correspond to much larger values of the cost and would
represent traps which would impede progress towards a global minimum.
Llacer et al (2003) have presented an extensive study in which the problem
of whether the presence of non-convex spaces affects the optimization of clinical
cases in any significant way is addressed. It has always been well known that the
use of a pure quadratic dose cost function with no constraints is a convex problem
with a single minimum. Work by Deasy and by Bortfeld independently in the
1990s showed that if dose-volume constraints were applied, then the problem
becomes non-convex and potentially then contains local minima. As is well
known, techniques exist to overcome the problems of local minima and achieve
global minima. The study by Llacer et al (2003) is seminal because it presents
techniques whereby the behaviour of beam-space, dose-space and cost-space
very close to global minima are systematically explored with the use of doubleprecision algebra and very small thresholds for determining the change that leads
to convergence. A very large number of clinical cases was examined and it was
found that either there was no evidence of multiple minima or that, when there is
evidence of multiple minima, the minima are so close together in cost space that
they do not affect the resulting dose distributions of the optimization although
the resulting fluence maps can be quite different from each other. Fourteen
different clinical cases were studied. It was expected that the results would largely
be independent of optimization technique and a Newton-gradient technique was
used. Zhang et al (2004b) came to similar conclusions and also claimed that the
Newton technique was the fastest of all inverse-planning methods and robust even
in the presence of local minima.
Inverse problems in IMRT are strongly degenerate in cost due to ill
conditioning. It is plausible that optimization is attempting to reach a very
broad minimum with extremely small gradients in the nearly flat final part of
the optimization. Similar discussions have been made by Alber et al (2002a,b).
Caccia et al (2001) have shown that advantage may be taken of the local
minima in inverse-planning algorithms. It was shown that the error due to the
implementation through MLCs is comparable with the differences introduced
by the rather wide distributions of solution, justifying simple and computionally
cheap approaches to the optimization process for typical cases.
Choi and Deasy (2002) have investigated whether inverse planning using the
gEUD concept leads to local minima. It was proved that when the only parameter
in the EUD ,a, is equal to or greater than 1 minimizing the EUD on a convex
feasibility space leads to a single minimum; if a is less than 1, then maximizing
266
EUD on a convex feasibility space leads to a single minimum. Jeraj et al (2002b)

have also shown that, with of variety of other objective functions, there are no
local minima in inverse treatment-planning calculations.
6.3.13 Sampling the dose matrix for IMRT optimization speed-up
All IMRT inverse planning requires the specification and use of a matrix which
links the dose to each voxel to the (unit) fluence in each bixel. Problems can
arise. Because photon scatter leads to dose many centimetres from the line-ofsight of a bixel, it is necessary to compute very large dose matrices which is
time consuming. Storing them also requires a very large space (or may even be
impossible leading to the requirement to calculate these matrices on the fly many
times during inverse planning). Even if they can be stored, looking them up and
repeatedly accessing all the elements of the matrix is time consuming. Hence,
one technique often adopted has been to truncate the effect of a bixel at a certain
lateral distance from the central axis of the line of sight. It is for this reason that
many inverse-planning algorithms are presented, in their test phases, using simply
the line-of-sight connection and some simple exponential dose model.
Thieke et al (2001, 2002) have proposed a technique whereby the large-scale
dose matrix is stored in full within a smallish radius (say about 25 mm) of the
central axis but beyond this it is sampled. To retain the total integral dose, the
sampled values are renormalized by the inverse of the sampling probability. This
(i) reduces the size of the stored matrix, (ii) speeds up access and therefore the
overall computation time and (iii) approximates the use of the full dose matrix.
Thieke et al (2002) explored the use of three probability sampling schemes. It was
then shown that the use of the dose-dependent probability-sampling scheme gave
results which are clinically no different from the use of the fully sampled dose
matrix, yet at about one-third the computation time (figure 6.19). Conversely,
the use of a truncated dose matrix led to significantly different results for some
clinical cases. The reason for the success of the probability sampling is that
whilst noise is introduced for individual beam elements, the superposition of such
elemental contributions smooths out the noise in the final dose distribution.
Cho and Phillips (2001) have presented alternative techniques to reduce
the computational dimensionality in inverse planning by using sparse-matrix
operations (figure 6.20). In the equation d = A b , A is an M N dose
calculation matrix, b is a column vector containing the individual beam element
weights and d is a dose vector whose elements are the dose sampling points
distributed in the 3D patient space. The matrix A can become very large.
However, only elements along principal diagonals relate to primary contribution
from individual bixels to individual dose points in the line of sight of those bixels.
All the other elements in the matrix A represent scattered radiation. Cho and
Phillips (2001) truncated all the elements below 0.03% of the maximum to create
an extremely sparse matrix. They then stored this matrix not as an M N matrix
but by storing the non-zero elements in one array and the locations, column-
267
Figure 6.19. Transverse CT slices of the head-and-neck case. The outlines of the
horseshoe-formed target and the spinal cord are shown as bold lines. The isodoses of
the dose distribution were optimized: (a) by a full matrix, (b) by a sampled matrix and (c)
by a cut-off matrix and are labelled with the absolute dose values in Gy. Additionally, the
directions of the seven beams are also shown. The 30 Gy isodose inside the spinal cord is
almost identical for the full and sampled matrix optimizations but is located more centrally
for the cut-off matrix optimizations. The use of a sampled dose-calculation matrix leads to
a significant speed gain for the dose calculation. (From Thieke et al 2002.)
wise and row-wise of these elements into other arrays. Typically they found
that this reduced the storage requirement for matrix A from about 1.4 Gbytes
to 94.4 Mbytes, i.e. to about 10% of the original size of the matrix A.
The effect of disregarding these small scatter components was then
investigated by repeating the optimization first with the sparse matrix and
secondly with the (normal) dense dose-calculation matrix. It was found that
the results of the optimization using the dense matrix and the sparse matrix,
when viewed as dose-volume histograms, showed that a comparison revealed
very little difference in quality of optimization. This observation fits in with the
observation of Webb (1991b) who also observed that when scatter was added to
inverse-planning code the overall features of the IMBs and the corresponding dose
distributions were largely unchanged.
Cho and Phillips (2001) provided the appropriate sparse matrix vector
multiplication code. The basis for this work was the observation that if, in order
to reduce the storage capability of a computer to store matrix A, an alternative
of decreasing the dose sampling resolution was evoked, the inverse planning
would deteriorate. The sparse matrix storage scheme used alternatively is the
compressed row storage format. All optimizations were performed with the
technique of projections onto convex sets (POCS). A hexagonal sampling scheme
to replace the more normal rectangular sampling scheme was proposed. Cho et al
(2003) showed that the use of sparse sampling, variably dependent on an organ,
can improve the conditioning of the ill-conditioned inverse problem. Specifically
268
Figure 6.20. A comparison of inverse planning using three different dose-sampling

resolutions: 0.0625 cm, 0.125 cm, 0.25 cm. The beam element dimensions were
0.25 cm 1.0 cm. (From Cho and Phillips 2001.)
the spectral condition number of the matrix linking beam space to dose space is
considerably reduced.
Deasy and Wickerhauser (2001) proposed to use wavelet decomposition to
decrease the memory required to store 3D pencil-beam dose kernels for IMRT.
It was found that 10:1 compression can be achieved with less than 1% error
anywhere in the reconstructed dose distribution. Deasy and El Naqa (2003) have
proposed a gridding technique whereby points used to control optimization are
more densely packed in regions where high dose gradients are required.
The issue of the time taken to perform IMRT optimization involves
consideration of the computational time spent evaluating the objective function.
Lee et al (2002b) have shown that it is possible to use just a fraction of the
voxels forming the PTV and OARs and compute the objective function in just this
fraction. It turned out that the final dose distributions obtained with a randomly
sampled subset of voxels was comparable to that produced when a full set of
voxels was sampled but with great savings in computer time.
Otto et al (2003) have conversely argued for a very fine spatial resolution
for the point-spread kernel based on the use of film dosimetry for the accurate
determination of dose in IMRT planning.
269
6.3.14 Creating a uniform PTV dose in IMRT; cost tuning

It has often been observed that IMRT optimization creates non-uniform dose to
the PTV in order to spare the OARs. Vineberg et al (2002) set out to study whether
and how it is possible to maintain the homogeneity of dose whilst simultaneously
sparing OARs. This was done in the context of using the UMPLAN TPS together
with its optimization tools for both IMRT and CFRT. The clinical application was
optimization of plans for head-and-neck treatment with parotid sparing to reduce
xerostomia. Each patient was replanned using nine equispaced fields. Targets
were defined with the aid of MRI. All work was carried out on a cluster of
COMPAQ Alphastations running VMS. Fast simulated annealing was used for
the optimizations.
This planning system makes use of costlets which combined to form a cost
function. The costlets symbolize different objectives. Some of these are as
follows: a dose-based costlet to force dose to specific values at a point; a meandose costlet to force the mean dose in a structure to a specific value; a dosevolume-based costlet to force the volume of a structure greater than or less than
some dose value to a specific volume value. Generally each of these costlets
involves the difference between actual running value and desired value raised
to some power. The power is the key to tuning the optimization, high power giving
a costlet more importance. Additionally, weights can be used when combining the
costlets to form the overall cost function. The highest priority costlets have the
highest power .
Vineberg et al (2002) specified a whole range of dose homogeneity values
for the PTV and showed how these ranges can be achieved in practice by varying
the power of the costlets. For all of 120 tested plans, the objectives were obtained.
Specifically, with respect to OAR sparing, it was found that subtle changes to the
costlet functions had effects on the trade-offs between PTV dose homogeneity
and OAR dose sparing. There was so much subtle variation however that no
specific rules of thumb emerged (figure 6.21). It could, however, be definitely
said that the PTV dose homogeneity could be kept within the traditional 5%
whilst simultaneously achieving all the objectives of sparing OARs. By relaxing
PTV dose homogeneity, better OAR sparing resulted. This implies that trade-off
discussion will become the dominant theme for the future of plan optimization.
Stated differently, the authors refuted the notion that target inhomogeneity is a
natural consequence of OAR sparing for head-and-neck cancer.
6.3.15 Importance factors

It should be remembered that IFs are just numbers that modify the effect of
contributing terms in a cost function. They have no physical dimensions or
meaning and sometimes large changes are required for them to effect small
changes in the dose plan (Bortfeld 2003).
270
Figure 6.21. A 3D view of the patient, target volume and the parotids and the nine IMBs
used for the plan including the intensity modulation for each beam. This is an example
plan used by the authors to attempt to obtain uniform PTV dose sparing OARs. It is
shown here, however, as a general example of the typical complexity of IMRT planning.
(Reprinted from Vineberg et al 2002 with copyright permission from Elsevier.)
Xia et al (2002b) have created an objective function for inverse planning

to include a power of the dose difference between calculation and prescription.
Then, by varying the power depending on the organ of interest, better plans with
more uniform dose to the target volume and/or more dose reduction to sensitive
structures can be obtained when compared with conventional quadratic objective
functions. Xia et al (2004) have shown that using a power of four instead the
conventional power of two led to improved results. The behaviour is not the same
as would be achieved by increasing the IFs for organs because increasing the IF
for an organ applies the same importance to all dose points whereas higher powers
271
give a greater contribution to the cost function from those dose points with a larger
departure from prescription.
Ayyangar et al (2003) have made a planning comparison of three systems for
inverse planning and found that direct use of the IFs set on one system (CORVUS)
to two others (ADAC PINNACLE and CMS FOCUS) did not yield suitable plans.
They thus confirmed that the selection of plan parameters is system dependent.
Yan et al (2003) used a fuzzy inference system, a form of artificial
intelligence, to arrive at a dose prescription and IFs for inverse planning for IMRT.
Yin et al (2004) have developed a fuzzy logic system for IMRT optimization in
which there are three parameters taken through a fuzzifier and a defuzzifier. These
parameters are the weighting factors for beams, the dose specification and the
dose prescription. The use of this form of artificial-intelligence-guided inverse
planning improves treatment outcome.
Phillips et al (2004) and Meyer et al (2004) have developed Bayesian
decision-making algorithms for optimizing IMRT. These provide the ability to
make use of an influence diagram to take account of as much information as is
known or can be estimated about the patient prior to inverse treatment planning.
Bedford and Webb (2003) have developed an inverse planning algorithm for
CFRT generating the most feasible solutions without the need to specify IFs a
priori. The system is known as AutoPlan.
6.3.15.1 Voxel-dependent IFs
Inverse-planning in its simplest form is driven by selecting some parameter
(e.g. the quadratic difference) that characterizes delivered dose compared with
prescribed dose. This concept is now well over ten years old. It is usual to assign
IFs to each region in space to allow the inversion optimization to do better in one
region than in some other region. In the past these IFs have been set globally for
the whole PTV and OARs. A disadvantage is that if a dose-volume histogram of
a structure fails to meet the objective, it is usually necessary to start all over again,
to re-specify the IFs and replan until the needed agreement ensues. Sometimes
this turns out also to be impossible to achieve because of the conflicts in the dose
specifications.
Xing et al (2000a) have shown that a two-stage optimization process is
required to arrive at IFs which allow the closest match to be achieved between
the expected and arrived-at dose-volume histograms. The optimization iteratively
cycles between optimizing a cost function in dose space and optimizing a cost
function in dose-volume-histogram space. Cotrutz (2002) and Cotrutz and Xing
(2002) have provided a very elegant solution (figure 6.22). At first a solution is
constructed based on a cost function in which IFs are globally specified to regions.
Then the dose-volume histogram is inspected noting its inadequacies. Then, using
a graphics tool, an inspection is made of all those voxels which specifically violate
the dose-volume histogram objective. The voxels are displayed graphically and
also flagged. Then the IFs of just these voxels are selectively adjusted and
272
Figure 6.22. Dose-volume histograms for plans optimized with unit value local importance
factors (the plain lines) versus optimized using higher value of local importance factors for
both the eye structures and the optic chiasm (the broken lines). Notice that by switching
on local importance factors that are non-uniform the protection of critical structures can be
improved. (From Cotrutz and Xing 2002.)
replanning commences. This opens up regions of the cost space for investigation
that have previously been unreachable. It was found that after a few cycles of
this process the dose-volume histograms are much improved (e.g. for OARs)
whereas those for the PTV are minimally disturbed. Essentially the improvement
is a consequence of introducing local level dosimetric constraints. Importances
are attached to voxels within structures in a differential way. Cotrutz and Xing
(2003a) showed, most importantly, that the same effect could not be obtained by
simply changing the overall structure-based IFs; it was the differential IF within
a structure that actually led to the observed improvement. Thus it was possible to
fine tune target and OAR doses flexibly.
Shou and Xing (2004) have observed that, in inverse planning for IMRT,
voxels within a structure are generally not equivalent in achieving their dosimetric
goals; this is because the spatial location of the voxel relative to the beams and
the other patient geometry affects the intrinsic ability of a voxel to meet its dose
prescription. This they describe as the dosimetric capability of a voxel and
a method is given to establish this parameter on a voxel-by-voxel basis. The
optimization strategy is then to assign a higher penalty to those voxels with lower
capabilities and this differential penalty scheme, when tried for a hypothetical
test case, shows that there is enlarged solution space when this non-uniform
importance is allowed. This enables one to break away from the mantra that, in
optimization, it is generally true that there is no net gain (i.e. an improvement in
one structure is usually accompanied by a dosimetrically adverse affect in another
273
structure). They comment that this form of local IF is another way of dealing with
attempting to a priori set a non-uniform dose prescription.
Jiang et al (2001a) have also highlighted the issue that not only the
magnitude of hot spots is important but also their location with respect to OARs.
This is because a hot spot close to an organ-at-risk might migrate into the organ-atrisk when patient motion is included. To take account of this effect, points in space
were spatially coded according to the distance of the points from the external
contour of the volume under consideration. In this way the spatial importance of
dose information can be factored into the inverse planning. The proof of principle
was demonstrated for a 2D prostate gland with five equally-spaced coplanar fields.
Deasy and Zakarian (2003) also applied dose-distance constraints, IFs that
varied within a structure according to the distance a voxel is from some defined
structure. For example, it was possible to weight those PTV voxels closest to
OARs differently from the general PTV weight.
6.3.16 Biological and physical optimization
De Gersem et al (2000) have developed a multileaf position optimization
algorithm which is based on a biophysical objective function. The tool iteratively
changes the individual leaf positions. The dose distributions before and after
adaptation are compared using this biophysical objective function. The method
has been applied clinically and ten patients were treated. For three prostate cases,
a better sparing of the rectum was obtained using leaf optimization. For seven
head-and-neck cases, a better sparing of critical organs was achieved as well as a
higher minimum target dose.
Holloway and Hoban (2001) have developed an IMRT inverse-planning
technique in which the cost function is a physically (dose)-based optimization
method but with IFs determined from biological characteristics. They argued that
the benefit of this dual scheme is the maintenance of homogeneous dose to the
target whilst simultaneously achieving the biological endpoints required.
Jones and Hoban (2002) have developed physically and radiobiologically
based optimization for IMRT. The method was developed in two dimensions and
applied to a plan with 35 beams, each consisting of 99 pencil beams. A ratio
technique was used to develop the fluences for each bixel. They compared a
ratio technique in which the ratio was specified using dose-based functions only
with two other techniques in which the ratio was determined using biological
parameters, EUD and the integral biological effective dose (IBED). The first
of these biological methods considered that the tumour comprised only tumour
cells and the second method assumed that the tumour comprised tumour and 10%
normal cells. The goal of the second biological method was therefore to spare the
normal cells within the tumour volume. It was found that the maximum tumour
dose using the biological optimization techniques could become extremely large
(greater than 100 Gy) particularly for the method which did not include normal
cells within the tumour volume. Including normal cells within the tumour
274
volume reduced maximum dose but still to nothing like the usual range of values
seen from the outcome of physically based optimization. Overall biological
optimization led to very inhomogeneous dose distributions because the EUD and
IBED parameters do not vary rapidly with dose. The authors also considered
the influence of the variation of the / ratio and itself. Altogether they
comment that the biological parameters are very poorly known as is the fraction
of normal cells in a tumour volume and, somewhat surprisingly, the last sentence
of their paper concludes that one should stick to physically-based optimization
techniques. This study represents a bold attempt to improve treatment planning
using biological modelling but at a state of knowledge where understanding of
the biological parameters is so poor that it is unlikely that workers would trust the
outcome of these biological optimizations.
Oelfke et al (2002) have created inverse planning within the KONRAD
treatment planning system based on EUD to include biological affects in the
inverse-planning process. Wiesmeyer and Beavis (2003) have developed a
technique to incorporate knowledge of biological function into an EUD-based
optimization in which the relative voxel intensity in a functional image is used to
provide increased weighting.
Stavrev et al (2001) have proposed a hybrid physico-biological approach to
treatment-plan optimization based on the constrained minimization of a biological
objective function. The concept was to introduce physical constraints that specify
the minimum and maximum target doses but define the objective to be minimized
in terms of the weighted sum of normal tissue complication probabilities.
Vineberg et al (2001) have compared dose-volume and biologically-based
cost functions for IMRT plan optimization. They came to the conclusion that the
basis of the cost function is not the determining factor for the conformality of the
dose distribution. Rather it is the balancing of the power and weights between
various components of the cost function that drives the dose distribution to the
desired solutions.
Zackrisson et al (2000) have shown that, if the time for each fraction
increases with IMRT, it may be better to change the dose per fraction. They
have studied a radiobiological model of this effect emphasizing the requirement
to consider the fraction time when moving from conventional treatment with a
small number of beams to fully optimized IMRT.
Nahum (2003) has introduced a reminder that uncertainty in the and
radiobiological parameters can have significant impact on the calculation of the
TCP from inhomogeneous radiation in the PTV. Also is questioned the concept of
whether delivering a boost dose to part of the PTV non-simultaneously with the
other dose has radiobiological implications. Measurement of hypoxia could guide
dose painting. Dasu et al (2003) re-emphasized that TCP models require the use
of a distribution of radiosensitivity parameters in order to obtain believable values
for clonogenic cell density and for radiosensitivity when such models are fitted to
clinical data (as discussed by Webb and Nahum 1993).
275
Wong et al (2002b) have described a small-animal radiation research

platform. This comprises a rotating ring gantry with a 60 cm aperture and three
225 kVp x-ray beams and a 20 cm 20 cm flat-panel amorphous silicon imager.
The goal is to reproduce as far as possible the conditions for human irradiation,
noting that generally when animals are irradiated in the laboratory it has been
with uniform radiation. The authors want to study in-vivo TCP and normal tissue
complication probability, interactions at the tumour and normal-tissue interface
and the quantitative study of radiation-induced gene therapy. Pelizzari (2003)
has also commented on the possibility to influence gene therapy using highly
conformal dose distributions.
6.3.17 Pareto optimal IMRT
Schlegel et al (2003), Bortfeld (2003) and Thieke et al (2003a, b) have described
how KONRAD creates a database of potential treatment plans which cover the
space of Pareto-optimal plans (figure 6.23). These are the plans in which an
improvement in one organ will inevitably lead to a worse result in at least one
of the other organs. Then after the automatic generation of this database, the
treatment planner uses an iterative search tool to select the optimal plan for the
considered case. Kuefer et al (2003) emphasize the advantages of this approach.
(Vilfredo Pareto [18481923] was a socioeconomist who stated the principle that
one cannot make any one person better off without making someone else worse
off.)
6.3.18 Combined CFRT and IMRT
Gaede et al (2001) addressed the familiar problem that IMRT inverse planning
generates fluence modulations that significantly depart from conventional
treatment beams. For three-field planning it was decided to constrain one field
to be uniform and modulate only the other two. This provides a solution which
is closer to conventional experience but nevertheless allows some modulation.
They also, somewhat unusually, chose parallel-opposed modulated fields with
the modulation identical for each pair of such fields. This contradicts the usual
advice not to use parallel-opposed fields for IMRT. Gaede et al (2002, 2004) have
argued that beam directions for an N-beam plan are likely to be favourable beam
directions in the search for a suitable (N + 1)-beam plan. So they addressed the
issue of how many beams are required for IMRT in this systematic way until the
addition of further beams does not further reduce a cost function. At each stage
of adding a beam, full inverse plan optimization was performed.
Vervoort et al (2001) have included two orthogonal static fields with five
IMRT fields for a 3D IMRT treatment plan. It is these two orthogonal static fields
which can then be verified using portal imaging on each treatment fraction to
ensure accurate patient positioning.
276
Figure 6.23. The figure shows the exploration of the Pareto front for a head-and-neck case
by brute force methods. Each black dot represents one treatment plan and 256 plans are so
represented. The fat dots represent the Pareto front for this case, being the set of efficient
treatment plans. Pareto optimization involves making an improvement in one particular
parameter, for example that shown on the x-axis, which corresponds to a worsening of the
parameters shown on the y-axis. (From Bortfeld 2003.)
Al-Ghazi et al (2002) have compared IMRT with conformal treatment with

a combination of the two. They found that, in some circumstances, the combined
therapy was superior to either of the two separate therapies alone.
Coolens et al (2003) also showed that the conformality of plans did not
dramatically deteriorate when a few beams were constrained unmodulated in a
five-field plan.
6.3.19 Split modulation
Grosser et al (2002) have developed a procedure for minimizing the number of
field components in IMRT using the step-and-shoot approach. Compared with
leaf sweep, the minimal field number is up to 50% lower for intensity maps
calculated with KONRAD.
For IMRT, Webb (2002c) described a simple technique which splits any
N N matrix I of fluences into two mathematically equivalent matrices IA
277
Figure 6.24. The upper panel shows a 15 15 IMB matrix to be split into two. The matrix
is random with a peak value of 10. The lower two panels show the result of applying the
algorithm of Webb (2002c).
and IB summing to 2I , each of which requires considerably fewer segments to

deliver IMRT using the jaws-plus-mask technique (Webb 2002b) (figure 6.24).
When IA and IB are delivered alternately at fractions throughout the treatment
the total number of field segments required is between one-half and two-thirds
that of delivering matrix I when unadjusted.
Faria et al (2002) presented clinical evidence to show that, when a 60 Co unit
was used to treat a four-field plan but only two fields were treated each alternate
day, the results were no different in terms of survival and late side effects to the
conventional delivery technique of delivering four fields each day. This is of
relevance to the recent suggestions from Grosser (2000), Grosser et al (2002) and
Webb (2002c).
278
6.3.20 Summary on inverse-planning techniques

Inverse planning, first proposed by Brahme (1988), has had an explosive
development in the last 15 years or so and there seems no end to the flow of
papers on this topic. Gradient-descent and simulated-annealing techniques have
remained popular. Algorithms based on EUD, genetic algorithms and maximumentropy algorithms have been developed. Attention has focused on the form and
basis of cost functions, both physical and biological. There has been continued
controversy over the existence of local minima in such cost functions. Techniques
have been developed to speed up inversions through clever sampling schemes.
IFs have been defined on ever smaller spatial scales rather than being applied to
whole targets or organs. Some studies have addressed the geometrical factors
that lead to specific planning outcomes (gradients and dose uniformities). Inverse
planning is now standard practice inside commercial planning systems and there
have been some comparisons of commercial systems and of planning algorithms
in general. Sadly it is next to impossible to make clear consensus statements.
Inverse-planning algorithms like cars have their good and bad points, their fans
and critics and it may be that one size does not fit all.
6.4 New forward-planning methods for IMRT; direct

aperture optimization
Generally IMRT plans are computed by inverse-planning techniques. These
include iterative techniques, algebraic techniques and a whole gamut of
sophisticated algorithms (see section 6.3) which can include dose-volume
constraints and individual weightings for different tissue structures and even
voxels within them. They form the basis for most of the IMRT planning taking
place in todays commercial treatment-planning systems. These commercial
developments go back to the early work done by, for example, Brahme (1988),
Bortfeld et al (1990) and Webb (1989, 1991a, b, 1992).
Siochi (2000b) has pointed out that simpler IMRT may be possible and,
in many circumstances, effective by performing a field-within-field technique.
A few fields with a few intensity levels may lead to sufficient conformation.
For example, Vanregemorter et al (2000) have used the ADAC PINNACLE3
treatment-planning system to do forward planning for IMRT. Fields consist
of four or five consecutive identical-gantry-angle fields with different apertures.
These shapes are determined to conform to different dose levels within the patient.
The relative weights of the fields were optimized by the planning system in order
to yield equal dose over identified points within the PTV. It was concluded that the
methodology of forward IMRT planning was a reasonable alternative to inverse
planning. The breast IMRT technique at the William Beaumont Rose Cancer
Center is another example of a field-within-field method (see section 5.7.1 and
Kestin et al 2000a, b). Beaulieu et al (2004) have discussed ways to create such
apertures automatically.
New forward-planning methods for IMRT
279
6.4.1 Segmental inverse planning at Thomas Jefferson University (TJU)

In recent years the group at Thomas Jefferson Medical College in Philadelphia
have challenged the conventional wisdom that inverse-planning techniques are
always required for IMRT. Instead, they have developed a form of forward
treatment planning for, what might be called, simplified IMRT or segmental
IMRT. The technique is to determine a beams-eye-view field for the complete
PTV and then to select sub-fields of different and smaller areas, each collimated
by a different pattern of the MLC, to irradiate those parts of a PTV not in line
with OARs. The Cimmino algorithm can then be used to optimize the weights of
these fields having specified dose constraints and IFs for the various tissues. Xiao
et al (2002a) have presented other mathematics of optimizing the intensities of
segments in an aperture-based inverse-planning technique. They showed that the
solutions use the least intensity and are very similar to those generated by the
well-known Cimmino algorithm. Chen et al (2002b) elaborate on aperture-based
optimization.
This approach is, of course, controversial because it somewhat kicks against
many conventional developments during the last decade and Xiao et al (2000a)
were careful to point out that the method probably works best when the target
structures are moderately well separated from the OARs and/or have a fairly
convex shape. They pointed out that the more concave a PTV becomes, the
more likely it is that full inverse-planning techniques are required. They are
certainly not rejecting these. Indeed, they have used the CORVUS system for
routine inverse treatment planning. This technique falls into a class of techniques
called segmental IMRT. Similar techniques are being developed at Ann Arbor
and in UCSF, particularly for irradiation of dominant intraprostatic lesions. The
paper by Webb (1991a) had diagrams of the main concept. Verhey (2001) has also
emphasized the importance of studying simple two-segment IMRT compared to
full IMRT.
Galvin et al (2001) use, for segmental planning, a combination of forward
and inverse planning to derive IMRT fields which are ten times more efficient to
deliver than those consequent on the use of simulated anealing via the NOMOS
CORVUS system. This DAO has been implemented inside a CMS treatmentplanning system. They have compared the outcome with an equivalent plan
produced using CORVUS which gives very noisy data and have credentialled
inverse planning by dose-volume histograms. It was shown that in one case
the MUs for a CORVUS MIMiC plan were 3552, for a CORVUS dMLC plan
were 864 and for the DAO plus MLC plan were only 296. It was found that the
number of segments was 478 for the CORVUS plan and only 49 for the DAO
plus MLC plan and thus concluded that small-field dosimetry was avoided and
that MU efficiency improved. Xiao et al (2000a, b), in this group, compared the
outcome of the forward-planning approach with a corresponding plan computed
using the NOMOS CORVUS treatment-planning system and their study showed
one particular treatment-planning case in which the dose to the rectum could
280
be considerably spared using the forward treatment-planning technique without

greatly compromising the dose homogeneity in the target. This Thomas Jefferson
University approach was also implemented inside the Elekta PrecisePLAN IMRT
system. Simplified accounts of this work appeared in Wavelength (2000a, 2003c)
emphasizing that this planning is an extension of conventional planning.
Galvin et al (2000a) have applied the segmented-field IMRT technique to
the breast. Selected isodose curves were projected onto the plane perpendicular
to the tangential fields and these were used to define and weight additional field
segments to improve the homogeneity of dose to the breast. Galvin et al (2000b)
have stressed the advantages of the forward-planning technique. These are that
it does not require a translator or interpreter, the intensity patterns are simple
and quite intuitive, QA is simplified because small fields are avoided and each
gantry angle includes at least a segment that conforms to the beams-eye view
of the whole target and standard verification port filming can be used. The
method can be applied without the need for any additional hardware or software.
The segments are selected based on rules that are a modification of those used
for traditional forward planning. Chen and Galvin (2001) have shown that a
useful technique for forward-planning IMRT is to add one new field on each
iteration, renormalizing the beamweights of the other existing field components,
so that the plan is improved. This avoids numerous difficulties encountered in
beamlet inverse planning. Chen et al (2001b) have refined the automatic-apertureselection technique for IMRT plan optimization. Segments are sequentially and
continually added to the calculation with re-normalization of the other beam
weights until an optimum set (usually less than 100) of segments is obtained.
Bednarz et al (2002) in this group have created an optimization technique
based on predetermined field segments. They considered nine fields with
several segments per field. Then they optimized the weights of these using a
mixed-integer programming technique to satisfy dose-volume constraints. They
compared the outcome to the optimization of field segments based on the
Cimmino algorithm and also compared to a full inverse-planned technique using
CORVUS. They concluded that the segmental optimization techniques could
perform as well as full modulated-bixel IMRT with the additional advantage
of simplicity, of easier QA and of stability. Earl et al (2003c) provide similar
evidence for the reduction in the number of segments.
6.4.2 Aperture-based planning at the University of Ghent
The group at the University of Ghent have pioneered the use of aperture-based
IMRT (see review by Webb [2000d] and reports on the developments in IMRT
at the University of Ghent to that date [Wavelength 2000a]). One patient was
treated in 1996, five patients in 1997, 17 patients in 1998, 47 patients in 1999
and 68 patients in 2000. By May 2001 the number of patients stood at 175. The
delivery technique gradually evolved. Up until May 1997, the MLC was operated
manually. Between then and 2000, the Elekta dynamic prototype equipment was
281
used and by 2001 Elekta RTDesktop was in use. A major application has been
to paranasal sinus cancer. Planning is performed using the GRATIS treatmentplanning system on a DEC Alpha computer which in 2001 had an interface
to two Elekta accelerators, an SL25 and SL18 both then operating the Elekta
Javelin operating system with MLCi and IMRT 3.1.0. Alternatively the Elekta
RTDesktop system can be driven. GRATIS is linked to both systems through
software based on the DICOM standard (Wavelength 2001b). Claus et al (2000)
have shown that the anatomy-based IMRT approach at the University of Ghent
does not necessarily lead to a large increase in MUs compared to conventional
radiotherapy.
De Gersem et al (2001a) have described in detail the anatomy-based
segmentation tool (ABST) This generates superposed segments which, after
beamweight optimization, results in IMBs. ABST is a complex algorithm but
essentially grows segment shapes that are in accordance with principles elucidated
over 20 years ago by Brahme et al (1982). For example, the beamweight of
radiation needs to increase sharply close to the canyons created by the views of
an OAR in order that the PTV be uniformly irradiated. This is implemented
algebraically in the technique and a feature of the technique is the automatic
generation of multileaf collimated field shapes. These are not created a posteriori
after intensity-modulated inverse planning. Also the geometric constraints of the
MLC can be included. This technique was applied to specific head-and-neck
treatments (figure 6.25).
De Gersem et al (2001b) have described the details of a tool for adjusting the
segment shapes and weights for step-and-shoot IMRT. The tool is called SOWAT
(Segment Outline and Weight Adapting Tool). The way this works is as follows.
Initially beam segments are created using geometrical views of target and OAR
structures. These are then optimized for beamweight to create conformal therapy.
The tool SOWAT then starts with the existing set of geometrical shapes and the
corresponding dose distribution and adjusts the shapes sequentially and cyclically
to attempt to improve the distribution (figure 6.26). It does this by inspecting
each shape, first making the adjustment with quite large leaf position changes,
cycling around and accepting changes which are beneficial and gradually reducing
the leaf-position change at each iterative cycle until a better set of segments is
generated. In an example, typically 41 cycles were executed.
The dose-computation engine inside this technique is much simpler, for
computation speed, than that which is used to finally compute the dose from
geometrical shapes. Once the final set of shapes is found, the ADAC PINNACLE
treatment-planning system or the differential scatter-air-ratio method of GRATIS
are used for a final dose calculation. The GRATIS system is also linked to the
ADAC PINNACLE system for image fusion and contour delineation and to the
PEREGRINE dose-planning Monte Carlo module and also linked to the HELAX
planning system. CRASH and SOWAT home-made software developments are
required because there are no equivalents in commercial systems (Wavelength
2001b). De Gersem et al (2001b) showed the technique applied to two particular
282
Figure 6.25. An example of the final leaf-and-jaw positions for all segments created by
ABST. (Reprinted from de Gersem et al 2001a with copyright permission from Elsevier.)
planning problems: a two-phase plan for elective nodal irradiation on both sides
of the neck and an ethmoid sinus cancer case.
Claus et al (2001) have performed a statistical evaluation of the SOWAT leafposition optimization tool for IMRT of head-and-neck cancer. This tool evaluates
the effects of changing the position of each collimating leaf of all segments on the
value of an objective function only retaining changes that improve the value of
the objective function. Although clinical implementation of IMRT at the Ghent
University Hospital started in 1996, before December 1999 IMRT plans were
283
Figure 6.26. Dose distribution of the first phase (046 Gy) of the treatment plan in a
transverse slice through the body of C2. Speckled areas are enclosed by contours of lymph
node regions II left and right, which are part of a CTV1. The light grey areas contoured in
white are the left and right parotid glands. (a) Dose distribution before the use of SOWAT,
(b) dose distribution after the use of SOWAT. (Reprinted from de Gersem et al 2001b with
generated without the SOWAT tool. After December 1999, SOWAT was in use.
Thirty head-and-neck patients were treated with IMRT between December 1999
and January 2002. Two distinct patient groups were distinguished: pharyngeal
and laryngeal tumours (17 patients) and sinonasal tumours (30 patients). Dose
statistics of the treatment plans with and without SOWAT were analysed.
It was found that, when using SOWAT for the pharyngeal and laryngeal
cases, the PTV dose homogeneity increased with a median of 11%, whilst the
maximum dose to the spinal cord was decreased for 14 of the 17 patients. In
four plans, where parotid function preservation was a goal, the parotid mean dose
was lower in one plan without SOWAT and in four plans with SOWAT. For the
sinonasal tumours, the PTV dose homogeneity increased with a median of 7%
and SOWAT lowered the mean dose to 53 of the 63 optic pathway structures. It
was concluded that SOWAT is a powerful tool to perform the final optimization
of IMRT plans without increasing the complexity of the plan or the delivery time.
284
Figure 6.27. Showing how three direct aperture shapes, each of which meets the
constraints imposed by the Elekta MLC, combine to give an intensity map for one beam
direction in which six non-zero beam intensities occur. (From Shepard et al 2002b.)
6.4.3 Direct aperture optimization (DAO) at the University of Maryland

Shepard et al (2001, 2002b) have described the DAO technique for step-and-shoot
IMRT. At each stage of an iterative optimization, both the shape and beamweight
of a segment can be adjusted. Initially all shapes are set to the beams-eye
view of the PTV. Then shapes are adjusted by randomly selecting MLC leaf
pairs. By this means the number of segments can be strictly controlled. All
MLC leaf constraints can be built into the optimization. If desired, shapes can
be constrained to have a minimum size and/or minimum number of MUs thus
overcoming concerns about the use of small shapes and low MUs. This approach
contrasts vividly with that of first finding IMBs and then a posteriori sequencing
them. Simulated annealing is used to minimize a quadratic cost function but with
dose-volume constraints inbuilt (figure 6.27).
Because of the possibility of overlapping apertures the number of fluence
levels per orientation is N = 2n 1 where n is the number of apertures. Thus
285
very few apertures yield many modulation levels. With conventional two-stage
sequencing, it is the other way round. It is claimed that DAO can reduce the
requirement to only three to five apertures per beam direction. Jiang et al (2004b),
for example, showed that, for a seven-field prostate plan, no more than seven
apertures were required since the objective function did not improve with more.
This corresponds to 27 1 = 127 fluence levels, rather more than suggested by
earlier workers such as Stein et al (1997).
Shepard et al (2001, 2002b) performed dose calculation using Monte Carlo
EGS4 with pencil beams computed for all needed options. The final configuration
was also recomputed. The method was applied to a phantom, a prostate patient
and a brain patient and computational runs were made to study the increasing
conformality and rate of increase as more beam segments are added. The
successful avoidance of cost function local minima was also investigated by
rerunning many cases with different random number sets.
Plan parameters were compared with corresponding CORVUS-generated
plans which use the two-stage process. It was found that both the number of
segments and the number of MUs were greatly (by about a factor of five) reduced
for the DAO method (Shepard et al 2003a).
The DAO method clearly obviates the whole problem of finding segments a
posteriori from modulated beams. Yu, at the AAPM Summer School (Palta and
Mackie 2003) described this as having the solution before one has the problem.
A strength is that it does away with bixel-based optimization; a weakness is that
it only does this for MSF (not the dynamic) MLC delivery (Shepard et al 2003a).
DAO also allows the concept of angular cost to be introduced. This is the cost
of omitting a beam angle and the philosophy is that where the angular cost is
low, this is a region in space where it is not necessary to develop large numbers
of different field shapes. It might be useful to constrain the variation of shapes
between angles which are close by in space.
Earl et al (2002) have simplified IMRT planning with DAO. Two to
five apertures per beam were sufficient to produce treatment plans that were
comparable to those produced with other inverse treatment-planning systems.
Earl et al (2003a) have presented an inverse-planning technique for IMAT based
on DAO. The two independent parameters to determine are the field shapes and
the value of the beamweight at each field shape. DAO starts by setting each
aperture equal to the projection of the PTV and the beamweights to one. Then
different apertures are arbitrarily selected and tested for whether they improve the
dose distribution. The technique is based on simulated annealing and iteratively
cycles until no further improvement is obtained. The width of the distribution
of the field size and the beamweights gradually reduces as iteration continues.
The technique is computationally demanding and so compression techniques are
used to reduce the amount of space required to store the pencil beams by a
factor of about 500. At all stages of iteration, the aperture shapes are checked
to determine if they satisfy the delivery constraints. The objective functions used
were a ranged-based and a dose-volume-histogram-based weighted least-squares
286
function. Several planning cases demonstrated the efficacy of the technique. Earl
et al (2003a) argue that the absence of a direct robust IMAT inverse-planning
technique has held up the clinical implementation of a technique that was first
announced in 1995 (Yu 1995). Earl et al (2003b) have described the use of the
DAO technique to create IMRT using only the jaws of an accelerator.
6.4.4 DAO wobbling the MLC leaf positions
Van Dalen et al (2000) observed that when a too mechanistic formula is used for
setting the MLC leaves for static therapy (e.g. adding a fixed margin to the beamseye view of the PTV), then the 95% contour does not always acceptably sheath the
PTV and spare nearby OARs. This is the case when multiple, and possibly noncoplanar, beams superpose. They developed a technique which iteratively adjusts
the individual positions of each leaf pair to optimized a cost function representing
the conformality. Ma et al (2000b) also developed a method to shape radiation
fields using dosimetric concepts.
6.4.5 Other forward-planning studies
Bos et al (2001) have compared two methods of creating beam segments
for static-field IMRT of prostate cancer. They studied one five-field prostate
case with a simultaneous boost. The two systems compared were a forwardplanning system in which beam segments were designed semi-automatically
within the UMPLAN treatment-planning system with subsequent optimization
of segment beamweights and the second system was the KONRAD treatmentplanning system which automatically generated fluence modulations which, when
segmented, again yielded static field components. The overall conclusion was
that inverse planning of segments for static IMRT did not improve the 3D dose
distribution compared to forward planning of segments. Bos et al (2004) extended
this study comparing forward plans created by UMPLAN with inverse plans
created by HYPERION and for five patients. It was found that inverse planning
created a more conformal plan, specifically reducing dose to the rectum but
that the inverse plan required typically about three times the number of field
segments. This study was a collaboration between the groups at The University
of Tubingen and that at the Netherlands Cancer Institute who routinely use
segmented irradiation for the prostate (see later).
Bar et al (2002, 2003) have compared forward and inverse planning for a
particular case where a treatment aimed to spare the function of parotid glands.
The forward technique constructed segments which viewed the tumour but
excluded views of the OAR and then weighted these with computer optimization.
The inverse-planning approach instead produced a fluence-modulated beam that
was then interpreted. In general, the resulting number of segments for the inverseplanned IMRT delivery was of the order 35 compared with just 11 for the forward
approach even though the same field directions were used. It was found that
287
the inverse planning had the potential to significantly improve dose distributions
compared to forward planning for a complex tumour sites in the head and neck.
The two-segments-per-field technique is also in use at the Netherlands
Cancer Institute. They also use HYPERION and have compared the forwardplanning technique that they have been using with the HYPERION inverse
planning. The variation in PTV between CT and MRI has been quantitated and
movement of the PTV has also been quantitated through the use of repeat CT
scans. The CT volumes were some 40% larger than the MR volumes.
Li et al (2002,2003a) have also proposed a DAO technique whereby the
beam directions and the number of apertures are defined ahead of a calculation
which then uses a genetic algorithm to determine aperture shapes. Cotrutz and
Xing (2003b) have also used a genetic algorithm to determine optimum aperture
shapes and beamweights by varying the MLC leaf positions (which map the
chromosomes). Petersen et al (2002) have also used forward planning to create
IMRT plans for 23 patients.
Corletto (2001) have compared inverse and forward planning for 1D and 2D
IMRT. It was concluded that 2D IMRT of the prostate led to better TCP of the
prostate than 1D IMRT which in turn was better than a standard conformal plan.
Lee et al (2004) have developed a technique of forward-planned IMRT
in which each beam direction has the beam divided into a number of discrete
segments whose weights are then optimized. These correspond to projections
of the critical structures. Over a period of seven years, 38 patients with headand-neck cancer had their treatments optimized by this method. The number of
segments was considerably lower than for inverse-planned IMRT without much
loss of conformality.
6.4.6 Summary on aperture-based IMRT
A review of IMRT planning written five years ago would probably have concluded
that IMRT absolutely requires inverse planning without exception. Certainly
techniques in which the bixel fluences are very finely spatially modulated do
require inverse planning for all the reasons stated when inverse planning was
first mooted. There are just too many variables to tie down by forward trialand-error techniques. However, recent years have seen the establishment of
IMRT techniques in which each field direction has only a few components.
Techniques have been developed to determine these shapes and their radiation
beamweights. Whilst this can certainly be done by inverse-planning methods,
alternative forward methods are possible. Many studies have addressed the
question of how the IMRT conformality changes as the number of such segments
is reduced. There is a trade-off between conformality and practicality. Maybe
some confusion surrounds this topic. Aperture-based planning can be inverse but
does not have to be. Also sometimes the technique is wrongly regarded as the
newest approach to IMRT whereas in fact it was one of the earliest proposed
before fully-modulated bixel-based IMRT was established (Webb 1991a).
288
6.5 Smoothing IMBs

6.5.1 Smoothing techniques from the Royal Marsden NHS Foundation
Trust
Many inverse-planning algorithms and commercial systems (e.g. CORVUS)
generate IMB profiles that have considerable structure. This is the desirable
outcome of the quest for high dose-space conformality. Webb et al (1998)
showed that the application of a median-window filter during the inverse-planning
iterative process could reduce the high-amplitude high-frequency bixel-to-bixel
intensity fluctuations without adversely damaging the corresponding deliverable
dose distributions. This reduced the number of MUs required for delivery with
consequent advantages on reduced leakage contribution to dose.
When CORVUS-generated profiles are realized experimentally using the
dMLC method of delivery, the MU efficiency can be quite small with unwanted
consequences. Also the interpretation of these fields leads to the generation
of small field segments, again with undesirable consequences. Webb (2001b)
showed that the features of beam space can be user-controlled by incorporating
beam smoothness into the optimization cost function to minimize these problems.
There is a trade-off between obtaining desirable features in beam space and high
conformality in dose space.
6.5.2 Smoothing techniques from the University of Virginia
Mohan et al (2000) have also investigated the influence of intensity fluctuations
on a number of issues in the delivery of IMRT by the dMLC technique. Firstly the
formalism on which the dMLC technique is based was restated, including giving
again the equations which determine the leaf patterns from the desired fluence
profiles. These closely follow the formalism of Spirou and Chui (1994, 1996).
The familiar MU-versus leaf-position diagram was constructed in terms of the
so-called effective fluence, being the combination of direct primary fluence and
an indirect fluence transmitted through, as well as scattered by, the leaves. By
assuming a constant leaf transmission, , leaf transmission can be accounted for
by the method of Spirou and Chui (1996). The so-called Stein equation shows
that the total number of MUs is given by the sum of a constant term representing
the time taken for the leading and trailing leaves to cross the field at constant
maximum speed together with a term representing the sum of the positive-fluence
increments along the track. From this follows directly the observation that, the
more complex the IMB becomes, the larger the number of MUs required to deliver
it. Six examples of increasingly more complex IMBs were shown to illustrate this
(figure 6.28).
The solution for the effective fluence cannot be accepted from the first pass
of the equations. This is because the position of the leaves changes the penumbra
effects and the MLC scattering. So Mohan et al (2000) adopted the familiar
technique suggested by Stein et al (1994) and iterated to convergence. Only a
Smoothing IMBs
289
Figure 6.28. A neat demonstration of the dependence of both the leaf trajectory and the
window width on the fluctuations of the intensity pattern. The left-hand pair of figures
shows the delivery of a flat distribution of fluence. The right-hand figure shows the delivery
of five deep valleys and the central figure the delivery of five medium valleys. As can
be seen, the number of MUs dramatically rises as the depth of the valleys increases. The
figures also show that the width of the open window would also dramatically differ between
these three different deliveries. (From Mohan et al 2000.)
few iterations were required. This slightly modifies the MU versus leaf-position
diagram. From these data the head scatter can be convolved (see, e.g. also the
work of Convery et al [2000]) to obtain the total fluence from which the 3D
dose distribution can be computed by folding in a dose-kernel and using the
linear attenuation coefficient for transmitted leakage. These calculations were
performed within the ADAC PINNACLE planning system.
Mohan et al (2000) used a clinical example (a nine-field prostate plan) to
show that, when dose-volume constraints on a number of OARs are applied, the
IMBs become more structured than they would be if these constraints were not
applied. Using some of the six model IMBs referred to earlier it was shown
that, the more complex the IMB is, the more likely it is that small field segments
will arise. There are a couple of consequences: (i) a large fraction of the
intensity at a point may be received indirectly and (ii) small fluence values may
be undeliverable due to the summations of leakages and scatter being larger than
the required fluence even when there is no direct contribution. Mohan et al (2000)
showed that by modelling the effects of leaf transmission, penumbra, head scatter
and MLC scatter, they can compute dose to within 1% of measured values.
290
They then suggested a filtering method (the use of a SavitzkyGolay filter)

which reduces IMB fluctuations. This has the desired effects of (i) reducing
significantly the number of MUs required for the delivery and (ii) creating a better
match between the desired fluence profile and the deliverable fluence profile.
Wu et al (2002c) have adopted a retro-mapping of IMRT beams to improve
delivery efficiency. Instead of filtering the whole intensities uniformly, the
proposed scheme adopts a strategy to filter only as much as needed to reduce
the MUs needed to deliver the beam.
6.5.3 Smoothing technique from the Memorial Sloan Kettering Cancer
Institute
Spirou et al (2000, 2001) have implemented techniques to generate IMB profiles
that are as smooth as possible. Two approaches were studied. In the first,
smoothing takes place at the end of each iteration. In the second, penalty terms
for the unsmoothness of the beam profile were included in the objective function
used to drive the optimization process. The latter method was preferred. It was
more efficient and required fewer iterations because the process of smoothing did
not undo the process of optimization as was the case in the former method. If
smoothing is introduced after each iteration, then the act of smoothing tends to
cancel out the act of conforming at the iterations and, after a while, the whole
dual process ceases to make progress. Also there is no inherent mechanism
for balancing the removal of genuinely required beam modulation from that
introduced due to noise in the iteration process. Conversely, when the smoothing
is controlled by the application of a term in the cost function, then it can be userconstrained. The user can balance how much aggression is aimed at conformality
versus how much importance is attached to smoothing. The form of the smoothing
is a SavitzkyGolay filter. Beam element fluences either side of the fluence bixel
being filtered are fitted to a polynomial. The order of this polynomial controls
the smoothness; the higher the order, the more peaks are preserved; the lower the
order the more effective is the removal of high-spatial-frequency modulations.
Second-order polynomials were preferred. The number of elements either side
can be user-selected. In general, symmetric choices were made. The technique
was tailored to the dMLC delivery with bixel size 2 mm along the direction of leaf
travel. Sharper dose gradients were obtained when the smoothing was included
in the cost function.
6.5.4 Smoothing technique from the University of Tubingen
Alber et al (2000) have developed the inverse treatment-planning system

HYPERION in which the limitations of the treatment equipment are specifically
taken into account for static and dMLC delivery with the result that the optimized
fluences are readily deliverable. The concept behind HYPERION (Alber 2001a)
is to provide a whole range of potential objectives for optimization. Essentially all
Smoothing IMBs
291
the experience of the doctor goes into dose-volume-histogram settings and what
you want is what you get. The basis of inverse planning is an optimization engine
which makes use of Monte Carlo dose data. Inclusion of biological and practical
constraints makes IMRT a more accessible and practical option. The shape of the
developed dose-volume-histograms is determined by the choice of constraints and
treatment intentions are made explicit. For example, it is possible to restrict the
dose variance to the PTV whilst simultaneously restricting the quadratic deviation
from the prescribed maximum dose. This is just one example from many. It
is possible to control the number of segments and to suppress undesirable field
shapes.
Smoothing is also built into the inverse planning and does not greatly effect
the dose distribution. Alber and Nusslin (2001) have described how to create
smoother IMRT fields to raise treatment-time efficiency by decreasing the number
of segments required under the constraints imposed by commercial MLCs. Bar et
al (2001c) have developed an inverse-planning algorithm to minimize the number
of segments for multisegment IMRT subject to dose constraints. This is achieved
by incorporating smoothing constraints into the fluence optimization algorithm
and also limiting the segments by a constraint on the size of the minimum size
of segment. Additionally, where possible, merging of segments is performed.
As a consequence, the complexity of the profile to be transformed and the
resulting number of segments is reduced, sometimes by as much as 60%. Alber
et al (2002a) have shown that the objective function in IMRT is entirely flat in
the neighbourhood of the minimizer in most directions. The dimension of the
subspace of vanishing curvature thus serves as a measure for the degeneracy
of the solution. This high degree of degeneracy is exploited to find solutions
which have more practicality (e.g. less complexity/more smoothness) than others
without degrading the treatment. They conjectured that it is the conflict between
the goals of dose in the PTV and sparing dose to adjacent OARs that leads to
high curvature in the objective functions and that once optimization iterations
have started to drive cost down this high-curvature part of the function, little is
achieved by more exploration of cost space. Conflict resolution spectroscopy is
discussed (see also Alber et al 2002b). Additional beams are only useful if they
aid resolution of these conflicts.
Laub et al (2000b) have studied IMRT of the rectum and specifically
studying small bowel toxicity. They have compared inverse plans created using a
commercial inverse-planning system with those produced by HYPERION. It was
found that IMRT could reduce the bowel volume irradiated to a dose of 95% of the
prescription dose by about 70% compared with conventional three-field treatment
techniques. However, whereas between 120160 step-and-shoot MLC segments
were necessary to deliver the five intensity profiles derived from the commercial
system, this number was dramatically reduced to between 20 and 40 segments for
the HYPERION plan. The reason this is possible is because HYPERION enables
the smoothness of the intensity profiles to be constrained.
292
6.5.5 Smoothing technique in the Nucletron PLATO TPS

The NUCLETRON PLATO inverse-treatment-planning system is being used at
the Christie Hospital, Manchester. The inverse-planning module has a medianwindow filter which can create smoother beams. It is possible to change the
width of the filter and apply it either to peaks or to troughs or both. The IMBs are
sequenced using a step-and-shoot interpreter developed in Utrecht (Wavelength
2001d). The general trend is that the higher the smoothing the fewer the number
of segments. Segments are then replanned onto an anthropomorphic phantom
for beam-intensity checks and plans are compared with film measurements.
Five-field clinical plans are also replanned onto a Rando phantom. These are
preliminary results requiring recalibration of films at right angles to the beam.
These class solutions are used for a hypofractionation trial for IMRT of the
prostate.
Mott et al (2001) have contributed to the debate over the relative merits
of inverse- and forward-treatment planning. They have developed smoothed
generic dose profiles, class solutions, applicable to specific patient and phantom
geometries and investigated how these class solutions then perform for cohorts
of patients. It was found that it was possible to apply phantom-created IMBs to
patients with additional compensation to get good conformality.
6.5.6 Smoothing technique at the Thomas Jefferson University
The Cimmino algorithm and mixed integer programming for inverse planning
have been compared showing that the Cimmino algorithm could produce very
smooth fluence patterns. Xiao et al (2001) have used the Cimmino algorithm to
segment multiple-static MLC fields in such a way that the number of segments
is minimized. At the same time this also leads to the generation of smooth
intensity patterns that have efficiency advantages. Xiao et al (2002b) have applied
Cimminos algorithm for solving the beamlet-based inverse optimization problem
for clinical cases and, in particular, have applied special filters in the optimization
process to improve the smoothness of the final fluence patterns. It was found
that the dose-volume histogram analysis showed similar dosimetric coverage of
tumour and sparing of nearby critical structures compared with plans made using
CORVUS but that the modulations were much smoother. Xiao et al (2003)
have shown that a projection method with dose-volume objective can also lead
to smoother beams.
6.5.7 Smoothing techniques at University of Maryland
Ma (2002a) has implemented smooth IMB sequences which accounted for the
hardware constraints. Large reductions (up to 90%) in the number of leaf
segments were found when applying the smoothing operation that minimized the
MLC leaf-travel distance, the total beam-on time and the number of leaf segments.
Ma (2002b) has extended his matrix decomposition method, which determines
Smoothing IMBs
293
leaf trajectories from IMBs to include the possibility to generate smooth beams
(figure 6.29). This exploits the multiple valid leaf-sequencing outcomes which
each and all correspond to a required IMB. The smooth beams so created require
fewer segments with all the advantages this carries. The technique was applied
to sets of random-pattern IMBs and also to clinical cases. The reduction in the
number of beam segments was greatest for physically unconstrained trajectories.
When interdigitation, TG and synchronization requirements were considered the
benefit fell. The effectiveness of the smoothing also depends on the complexity of
the IMBs. For complex cases with numerous treatment constraints, shallowness of
the cost function minimum may be too low to permit much room for manoeuver.
Wide-bottomed cost functions lend themselves best to the technique. Li et al
(2004a) further extended this work to include the possibility to eliminate leaf-end
abutments and simultaneously minimize the number of MUs and the number of
field segments (see section 3.1.6).
6.5.8 Smoothing techniques at University of California, San Francisco

Sun and Xia (2004) have implemented smoothing in IMRT planning in yet
another different way. Firstly they optimized a plan by a conventional simulated
annealing approach. Then they applied a smoothing filter to the IMBs. In doing
so, only those bixels labelled at projecting to specific structures (controlled by a
bixel structure index) were included in groups for smoothing. This smoothing
process degraded the conformality. So a third step was then taken to re-optimize
the weights of the segments from the outcome of the smoothing stage. They
showed for several clinical examples that this led to large reductions in the number
of components without violating the required conformality.
6.5.9 Smoothing techniques at Sichuan University, China

Hou et al (2004) have described a novel way to perform inverse planning that
not only improves spatial resolution but also increases beam smoothness leading
to the requirement for fewer segments. In this method, several (two and four
were tested) matrices of modulated bixels of conventional size 1 cm 1 cm were
produced for each beam angle with the N matrices displaced relative to each other
by multiples of 1/N cm (e.g. two matrices are spaced by 5 mm apart; 4 matrices
are spaced 0.25 mm, 0.5 mm and 0.75 cm apart.) Then the cost function was
specified so as to simultaneously optimize all these constituent matrices. The
resulting IMB was then the sum of the shifted matrices at each beam location. It
was found, using a test case and a clinical case of treating tonsil, that the resulting
matrices had higher spatial resolution, greater smoothness and yet simultaneously
better conformality was achieved.
294
Figure 6.29. Illustration of the algorithm of Ma (2002) for a typical IMB: (a) the tolerance
level at 0% that resulted in 42 segments, (b) the tolerance level at 5% that resulted in 34
segments, (c) the tolerance level at 10% that resulted in 25 segments, and (d) the tolerance
level at 15% that resulted in 12 segments. Notice the coarsening effect of the intensity
maps as the tolerance level increases. (From Ma 2002b.)
6.5.10 Summary on smoothing

When IMRT is delivered by the dMLC technique, the treatment time is directly
coupled to the variability of the IMBs through the Stein equations. Hence,
smoothing IMBs leads to shorter treatment times. When, alternatively, the
MSF technique is used, then smoother beams lead to the requirement for fewer
segments. Again this speeds treatment time although possibly now not so much
as in early days when the intersegment deadtime was a large fraction of the
treatment time. To some extent, this has obviated the need to reduce the number
of segments.
Incorporating MLC equipment constraints in inverse planning
295
6.6 Incorporating MLC equipment constraints in inverse

planning
Bortfeld (2000) has reviewed the step-and-shoot IMRT technique, concentrating
on the various hardware constraints and putting special emphasis on solutions that
reduce TG underdose and matchline effects. Leaf sequences have been developed
which keep the number of segments within practical limits. Step-and-shoot IMRT
has been compared with the dMLC IMRT technique and the advantages and
disadvantages of each method have been presented.
Alber (2001a, b) has also incorporated delivery constraints for static and
dMLC delivery into IMRT inverse planning and has discussed both hard (MLC
machine constraints) and soft constraints in IMRT treatment planning. Hard
constraints are built in to minimize the gap between opposing leaves and to
disallow interdigitation. Also some soft constraints are applied to minimize the
segment size, the segment width and the segment separation. At the University
of Tubingen, the inverse-planning system HYPERION is linked to the ADAC
PINNACLE planning system to drive the Elekta RTDesktop delivery system.
The aim is for seamless integration between the ADAC PINNACLE planning
system and HYPERION. CT data are taken first into ADAC PINNACLE for
contouring, then into HYPERION for inverse planning, then the segments back
into ADAC PINNACLE for dose verification and finally taken to the Elekta linac
for delivery of IMRT. HYPERION delivers step-and-shoot segments. Inverse
planning incorporates a step-and-shoot sequencer with variable fluence steps.
After optimization and sequencing, there is a re-optimization of segments because
the sequencing changes the dose distribution. The result is a set of weights
and segments, which do not violate the constraints. The idea of minimizing the
segment size is to improve the dosimetry and is very similar to one of the goals of
Webb (2001b). It is claimed that the dosimetry of the ADAC PINNACLE system
is accurate to 1.5% for 1 cm 1 cm fields.
Holmes (2001) has developed a technique for inverse planning in which the
leakage and head scatter are incorporated during each iteration of the optimization
process. The method has been applied to serial tomotherapy in just two
dimensions as delivered by the NOMOS MIMiC. This is to overcome the wellknown observation that, if the delivery constraints are not included in the inverseplanning stage, then when they are put in a posteriori they will deteriorate the
dose distribution, generally making it less conformal.
The way the delivery features are built into the inverse-planning system is
to identify active bixels as those which are in the beams-eye view of the PTV.
These active bixels may be open or shut for specific fractions of the total radiation
time. During the time in which they are closed, whose fraction f will be known,
these bixels will contribute to the active dose through leakage and this may be
factored in, knowing the transmission of the MIMiC vanes and the associated
back-up jaws. Correspondingly, the contribution from inactive bixels may also
be computed. The essential difference between this and other methods of inverse
296
planning are that these contributions, including head-scatter effects, are factored
in at each stage of the iteration during a steepest-descent method with a quadratic
least-squares objective function. The outcome will then be intensity-modulated
profiles which are already corrected for these physical effects. In Holmes study,
graphs of beamlet weights illustrate the effect of correcting the active beamlets for
leakage and head scatter and show the relative contributions. Dose conformality
improves with increased stratification of IMBs. This work has been extended
by Seco et al (2001) for inverse planning to create distributions which will be
delivered with the MSF and the dMLC techniques.
Constraints for segments shapes can be built into planning. The University of
Ghent have parametrized weirdness and put this into the objective function and
concluded that weird shapes were not needed. In particular, the field patterns
developed did not have the salt and pepper noise associated with CORVUS field
modulations.
Siebers et al (2002b) have incorporated MLC delivery constraints and
features (leaf transmission, TG effect) into an inverse-planning optimization code
at the Medical College of Virginia (figure 6.30). The algorithm is a gradientdescent technique. It was argued that the resulting plans conform more closely to
the plan constraints than plans in which the planning and delivery are considered
as separate operations. The integrated approach was compared with the sequential
approach for 17 patient cases, mainly but not exclusively in the head and neck.
In general, dose to the PTV was adequately realized in both approaches but
the integrated approach led to better critical organ sparing. It was observed
that if the sequential approach is used the dose-volume constraints have to be
overemphasized in order to achieve the actually required constraints. Also much
trial and error is required which is eliminated by the integrated approach. Because
the leaf sequencer tends to smooth profiles, the integrated approach leads to more
MU-efficient deliveries in all cases. This study refered to similar work by Cho and
Marks (2000) and by Holmes (2001). Lauterbach et al (2001) further reported on
the building of this sequencer inside the IMRT optimizer in order to ensure that
the fluence maps so created are deliverable. For a head-and-neck case and for a
prostate case, the resulting isodose distributions showed a clear improvement for
the deliverable optimization technique.
6.7 Beam direction optimization

Macklis et al (2000) have provided an editorial on IMRT treatment-planning
paradigms. They extended the definition of IMRT, from the simple indication
that photon fluence is varied spatially either through direct spatial variation or
through temporal variation, to include a new (so-called)world-view that IMRT
encompasses not only arranging for extremely high levels of target-structure
dose conformality but a very high degree of automated beam selection through
Beam direction optimization
297
Figure 6.30. Two flow diagrams indicating the difference between traditional IMRT
optimization and deliverable-based optimization. On the left is shown the flow diagram
for traditional IMRT optimization. Intensities are iteratively updated to improve the plan
quality until convergence. Following convergence, the optimized intensity is converted
to MLC leaf sequences and deliverable intensity distributions. On the right is shown the
flow diagram for the deliverable-based optimization method. Intensities are converted to
MLC leaf sequences and deliverable intensities inside of the iterative IMRT optimization
loop. The same leaf sequencer was used as in traditional optimization. (From Siebers et al
2002b.)
inverse treatment planning and a willingness to question the traditional teachings

of conventional radiotherapy.
The problems of determining the optimum beam orientation and the
optimum set of IMBs are coupled because the optimum IMBs will change
depending on the gantry and couch angles selected. There has been a considerable
amount of research trying to unravel the problems.
Holloway and Hoban (2002) have developed an IMRT planning technique
that starts with a very large number of beams. The process relies on gradually
reducing the number of beams used, based on the average fluence delivered by
each beam. Beams with the lowest average fluence are removed at each stage
of the iteration and the process is repeated until the desired number of beams is
reached. For the situation of 180 initial beams, gradually reduced by three beams
298
at a time, a plan was achieved vastly improved on that using evenly spaced beams
after ten iterations.
Gaboriaud et al (2000) have shown that use of IMRT with non-coplanar
beams improves the treatment of brain tumours. In their study, conformal
geometrically-shaped fields were compared with intensity-modulated fields for
the same beam directions, the greatest benefit being dose reduction in the OARs
especially the brain stem and the temporal lobes.
Woudstra and Storchi (2000) have developed a way to iteratively select both
the orientations and the beamweights of a set of beams which, in principle, can
be as large as 36 coplanar beams in [0,360]. The method does not use a planbased cost function but sequentially optimizes the placements of the next beam
by selecting that one which improves the plan most at that stage of selecting
beams. This is a no way back method as orientations and beamweights selected
early in the process cannot be deselected. However, by including a stage of
continuously varying the IFs, the method seems to work. Woudstra et al (2002)
have developed automatic techniques for selecting beam orientations and IMRT
beam profiles for pancreas, oesophagus, head and neck and lung and found that
the automatically generated plans, even with a few beam numbers (39), compare
well with equiangular beam set-ups.
Meedt et al (2001) have presented a new algorithm for optimizing beam
directions in IMRT. The algorithm starts by creating a standard N-beam treatment
plan using fluence-profile optimization. Then, one-by-one, new test rays with the
smallest desirable field size are added and optimized and then, following this
process, the beam which can best be compensated-for by the remaining beams is
removed and the next iteration is started. The process terminates when attempting
to add new beams takes the dose distribution no closer to the desired dose
distribution. Meedt et al (2002) have applied this beam-direction optimization
algorithm to head-and-neck IMRT. It was found that a small number (46) of noncoplanar beams can sometimes improve upon a larger number (1020) of coplanar
equally spaced beams. The algorithm employs a fast-exhaustive search of about
2000 directions equally spaced on the unit sphere to find a new beam best capable
of improving an arrangement and then deleting a corresponding less useful beam.
Meedt et al (2003) have developed a beam-direction search strategy for IMRT
based on selecting beams with paths of little resistance and an exchange scheme
to converge on an acceptable outcome.
Webb (1995) introduced the idea of creating a set of beams with maximum
geometrical separation to achieve a starting point for CFRT. This was done by
a Monte Carlo technique which maximized the sum of the angular separations
between beampairs. Wagner et al (2001) have capitalised on this as follows.
They have started with the previously described distributions. Then they rotate the
locked set to an orientation which works best. Then they individually reorientate
each beam within a preset small angular region. This is done to minimize a score
function which has the effect of avoiding couchgantry collisions, of avoiding
difficult entry orientations and of minimizing OAR dose. IFs were used to
299
ensure that avoiding more critical structures outweighed avoiding less critical
structures. When the individual beam directions are adjusted, they are done
so within a cone of 45 about their starting location. Clearly this removes the
precise maximally-spaced-out property of the isotropic beam bouquet. A clinical
case bears out the goal of the orientation adjustment showing the tuning of beam
directions improves OAR sparing and retains high dose gradients. The original
notion from Webb (1995) was never intended to be the complete clinical solution
so this advance has brought the notion to the clinical arena.
Levine and Braunstein (2002) have presented a theoretical exposition of the
number and directionality of beams required for 3D IMRT. The number of beams
predicted vastly exceeds current practicality. The paper is of theoretical interest
only.
Meyer et al (2001) have investigated an interesting and novel problem. When
N IMRT fields are equispaced, coplanar, around a PTV and when N is fairly large
(e.g. nine) some of the field arrangements can lead to intersections of the couch
and the beam which is not permitted. They studied this phenomenon for three
types of couch and came up with a set of rules to predict such collisions. They
predicted the percentage of times this would occur for each type of couch and
value of N for some model tumour geometries. Then they adjusted the angles of
beams so as to avoid the collision without seriously affecting dose conformality.
Meyer (2003) compared an equiangular seven-field plan with a beam-angleoptimized plan on an artificial phantom with two OARs and showed that the OAR
dose could be considerably decreased using the non-equispaced beams.
Das et al (2001a) have developed a new technique to automatically select
beam directions for IMRT when the modulations are relatively simple. The
technique comprises alternating optimization and elimination steps until the userspecified number of beams remains together with the appropriate modulation.
Das et al (2001b) have developed a beam-orientation scheme for IMRT based
on target EUD maximization. Essentially optimization oscillates between beamorientation and beamweight selection and, at alternate stages, those beams which
contribute least to maximizing the target EUD are sequentially removed until the
desired number of beams remains. Das et al (2003) showed that this EUD-based
technique yielded a satisfactory dose distribution for just three to five beams for
a prostate case compared with that produced by a large number of unselected
equispaced orientations.
Li and Yu (2001b) have also developed an IMRT algorithm that
simultaneously optimizes beam intensity and beam orientation.
A large number of approaches to beam-orientation optimization has been
made by Pugachev and colleagues. Here follows a detailed review of these in
date sequence. Pugachev et al (2000a) observed that the problems of optimizing
beam profiles and beam directions in IMRT are coupled. This problem was
solved by using a simulated annealing algorithm to determine the orientations
of nine beams in 2D IMRT planning. At each iteration and the angles selected,
filtered backprojection was used to compute beam profiles. This relatively fast
300
technique was practicable. It was shown that the resulting dose distributions
were improvements on those resulting from optimizing beam profiles at fixed,
regularly-spaced, gantry orientations. It was also shown that if simulated
annealing were used to also optimize beam profiles then the outcome was much
the same but took much longer to achieve. The work is analogous to that of
Rowbottom (1998).
Pugachev et al (2000c, 2001) have investigated the improvement of
conformality of IMRT with introducing non-coplanar beams, with and without
optimization of the beam-orientation direction. Baseline plans were created for
nine coplanar beams which were equispaced in the gantry-orientation angle. Dose
distributions were compared with the results of planning in which the beam
orientations were allowed to vary in a single plane and, alternatively, in which
beam orientations were constrained to vary non-coplanar inside a fixed domain
of gantry and couch angles. For each beam orientation tried, an intensitymodulated plan was created using PLUNC and the so called simultaneousiterative inverse-treatment-planning algorithm of Xing et al (1998). The beamorientation algorithm has a problem with local minima and so was attacked using
the technique of simulated annealing. The algorithm randomly selected beam
orientations, computed the beam profiles at such orientations and then accepted
or rejected the trial on the basis of a cost function. The simulated annealing
was controlled as usual by starting at a high temperature and then gradually
reducing the temperature to freeze the gantry and couch locations. The technique
is a brute-force technique, quite computationally intensive, needing upwards of
a 100 hr on a Silicon Graphics computer but yielded results. Pugachev et al
(2001) solved three problems: the first was a case of prostate cancer; the second
a nasopharynx case and the third a paraspinal case. It was observed that, for the
prostate cancer case, very little advantage accrued to optimizing beam directions,
whether coplanar or non-coplanar, in fact only a 5% improvement in cost function
resulted with very little extra sparing of bladder and rectum. However, for
the nasopharyngeal case, the cost function decreased by 27% for the plan with
optimized non-coplanar beam orientations and led to a significant sparing of brain
stem and optic chiasm. Correspondingly, for the paraspinal case, the cost function
decreased by 19% for the plan with optimized non-coplanar beams and there was
considerable sparing of the kidney and liver. It was thus concluded that the effect
of beam-orientation selection on the quality of the treatment plan varies from site
to site.
Pugachev and Xing (2001a) have decoupled the beam-orientation selection
from the fluence-modulation calculation for inverse planning for IMRT. The first
stage was to recognize and exclude bad gantry angles in which too much
radiation is delivered to OARs. The search space then concentrated on the
remaining good gantry angles and optimized the fluence profile using simulated
annealing. The results showed that the quality of the final treatment plan was not
affected by this restriction of the search space.
301
Figure 6.31. A pseudo beams-eye-view score as a function of beam orientation for

(a) a coplanar beam configuration and (b) a non-coplanar beam configuration indicating
preferential angles of attack towards the PTV. (Reprinted from Pugachev and Xing (2001c)
with copyright permission from Elsevier.)
Pugachev and Xing (2001c) have introduced a concept of pseudo beamseye-view (pBEV) to establish a framework for choosing beam orientations in
IMRT (figure 6.31). The idea was to introduce a score function for each
1 cm1 cm beamlet which crosses a target and to assign the maximum intensity to
that beamlet that could be used without exceeding the tolerance dose of the OAR.
The score function was then constructed in terms of these individual beamlet
functions. The idea then was to make a plot of the score function as a function
of orientation, decide how many beams to use and choose the directions which
have the largest value of score function whilst also ensuring that beams are not
antiparallel and are spaced out as widely as possible. It was shown that this
technique is especially useful for complicated clinical cases.
The pseudo beams-eye view is an extension of the classical beams-eyeview technique which just measures the quality of the beam direction as the binary
function based on whether the beam does or does not hit a sensitive structure. The
pseudo beams-eye-view technique accounts for beam modulation. The technique
was incorporated inside the PLUNC treatment planning system. Couch angles
were discretized in increments of 10 and gantry angles in increments of 5 .
302
The concept is based on observing that a beam orientation is preferable if it can

deliver more dose to the target without exceeding the tolerance of OARs or normal
tissue located on the path of the beam. For each beamlet crossing the target the
maximum intensity that could be used without exceeding the tolerance of the
OARs and generic normal tissue was calculated.
The technique starts assuming that only one beam is going to be used for
irradiation. Firstly, the voxels crossed by the beamlet were found. Secondly, the
beamlet was assigned an intensity that would deliver a dose equal to or higher
than the prescription dose in every target voxel. Thirdly, for each OAR or normal
tissue voxel which is crossed by the beamlet, the ratio was calculated by which the
beamlet intensity had to be reduced to ensure the tolerance dose is not exceeded.
The minimum ratio from that data was then identified. The beamlet intensity
was then reduced according to this ratio and the value obtained represents the
maximum usable intensity of the beamlet. These steps were then repeated for all
the relevant beamlets to obtain the maximum beam-intensity profile in which none
of the beamlet intensities can be further increased without violating the structure
tolerance. Then an overall score function, a single number, was calculated to
reduce these data to a single number.
It is emphasized that this maximum-intensity profile as previously defined is
not at all the optimum fluence profile for IMRT. It simply allows us to create
a plot of score function as a function of gantry orientation from which beam
directions can be chosen for subsequent optimization of IMRT. It was argued that
this captures the main features of the planners judgement of the goodness of a
beam direction. It is a trick for significantly reducing the size of the search space.
Once the plots of score function versus orientation had been determined, the
next task was to select a fixed number of beams and this was done by adding in the
separate criterion that beams should be angularly as-well-spaced-out-as-possible
and also not antiparallel.
Two clinical cases were investigated showing that the tool is an efficient
one for IMRT beam-orientation selection, acting as a filter. It certainly does
not provide the final beam configuration for IMRT treatment. This still requires
the intervention of a human or inverse planning. The fields so selected were
optimized using CORVUS and it was found that, for prostate therapy, five-field
techniques could be improved over the use of equally-spaced beam configurations
but that nine-field IMRT could not be so improved. However, when performing
treatment planning for paraspinal and nasopharangx in which the improvement
depended more on the degree of overlap with OARs it was found that, even
when nine fields were used, more suitable beam orientations could be selected
using the pseudo beams-eye-view optimization technique. It was argued that,
whenever a multidimensional dataset is projected onto a lower dimensional space
for decision-making by a human being, the details of the local quantities are often
buried.
Pugachev and Xing (2002b) have extended this work by developing a
technique that uses beams-eye-view dosimetrics (BEVD). The goal of the work
303
was to provide an efficient means to speed up the beam-orientation optimization

by incorporating a-priori geometric and dosimetric knowledge (figure 6.32). The
BEVD is a scalar scoring parameter that describes the ability of a particular beam
direction to contribute to the PTV dose independent of the contributions from
other beam directions. Having set a pre-set number of orientations, the BEVD
were computed for these orientations. Those with a very small value were then
rejected altogether from consideration. Thus the candidate orientations were prescreened. The BEVD was then used to weight the number of times a particular
beam orientation is sampled in an iterative sampling scheme, the logic being that
it is preferable to re-visit those orientations which are considered to be good.
The technique was implemented inside the PLUNC planning system. The
dose objective function was a weighted quadratic and the pre-ranked angular
search space was sampled using simulated annealing. Whilst, in principle, any
sampling density function could have been used, it was found that a sampling
proportional to the square of the BEVD score provided an efficient convergence
and speeded up the conventional simulated annealing calculation by about a
factor of 10. This is because the pre-screening and the weighting reduced the
chance for the algorithm to spend valuable computing time on those orientations
that were less likely to end up in the final solution. This improvement was
compared with the use of linear beams-eye-view volumetrics and found that the
two techniques gave the same gantry angles at the end of the calculation but that
the new technique was distinctly faster. Incorporating the prior knowledge also
improved the convergence behaviour of the simulated-annealing calculation based
on a realistic cooling scheme. However, the technique still took more than two
hours on a fast workstation. The whole technique is predicated on a fundamental
rule for estimation theory that the use of prior knowledge (Bayesian priors) will
lead to a more accurate estimation.
Pugachev and Xing (2001d) have developed a computer-assisted selection of
beam energy and orientation in IMRT. A score function was plotted as a function
of gantry angle for each energy, this being the largest intensity which could be
delivered from that angle without violating the tolerance of the structures located
on the path of the beam. An algorithm then selected beam angles and beam
energies from the two curves, such that the beam angles are as far apart as possible
and the better of the two beam energies at such angles is used. The technique has
considerable potential for simplifying IMRT treatment planning.
Pugachev and Xing (2002a) studied eight nasopharynx cases and ten prostate
cases to try to decide if class solutions were acceptable. They came to the
conclusion that they were not and that individualized beam configurations were
needed to maximally utilize the technical capacity of IMRT.
Djajaputra et al (2003) proposed a beam-direction optimization algorithm
for IMRT based on a fast dose calculation engine (table look-up method) and
a fast simulated annealing algorithm for beam-direction selection. This dosecalculation method is justified in that the matrix linking dose space to beam space
is very sparse and bixels are planned to contribute only to voxels in their line of
304
Figure 6.32. (a) The dose distribution of an IMRT para-spinal treatment with five
equiangular-spaced beams. (b) The dose distribution of the IMRT para-spinal treatment
with five beams obtained using the beams-eye-view-dosimetrics-guided beam orientation
algorithm. The incident beam orientations are indicated by arrows. (Reprinted from
Pugachev and Xing (2002b) with copyright permission from Elsevier.)
sight. For each selected set of beams, the fluence was optimized by a gradient
descent technique. The number of beams was fixed and not optimized. Several
planning cases were studied. For a three- or five- or seven-field prostate case,
beam-angle selection led to an improved plan compared with a uniformly spaced
set of beams. Rotating the uniform set did not improve these plans as far as
a beam-angle-optimized set, although the relative trade-offs in dose sparing to
OARs changed. For a uniformly spaced set with a large number of beams, the
Monte Carlo dose calculation
305
rotation had less effect. Similar behaviour was observed for a head-and-neck
case.
Generally, in IMRT planning, the optimization of beam orientations and
beam weights for CFRT become separate problems. Wang et al (2003a) have
combined the two problems and solved them using mixed integer programming
to optimize both the beam orientations and the beamweights simultaneously. The
problem was solved using four steps.
(i) First a pool of beam-orientation candidates was set-up with the consideration
of avoiding any patientgantry collisions and avoiding direct irradiation of
OARs with low tolerances.
(ii) Each beam-orientation candidate was represented with a binary variable and
each beamweight with a continuous variable (in this initial application each
beam could comprise one of four wedged dose distributions).
(iii) The optimization problem was set-up according to dose prescriptions and the
maximum number of allowed beam orientations.
(iv) The optimization problem was solved with a commercial ready-to-use
mixed integer linear programming (MILP) solver. After optimization,
the candidates with unity binary variables remained in the final beam
configuration.
Wang et al (2003a) use PLUNC to generate the dose and anatomical data
files and then a home-made program to write the MILP data files in the format
of MPS (a standard mathematical programming system). The time for set-up
depended on the number of fields and the same was true of solution and typically
both can be achieved in under an hour. It was then shown for two cases, the first
a prostate and the second a head-and-neck case, that the conformality of radiation
improved using this technique. The gains were much greater for the head-andneck case than they were for the prostate case. The application was anticipated to
be applicable to IMRT.
Schreibmann et al (2004c) have presented a technique which generates a
Pareto set of optimized solutions among which the user can choose.
6.8 Monte Carlo dose calculation

6.8.1 The debate over the usefulness of Monte Carlo dose-calculation
techniques
In the Medical Physics Point and Counterpoint, Mohan et al (2001a) have
debated whether or not Monte Carlo techniques should replace analytical methods
for estimating dose distributions in radiotherapy treatment planning. Mohan
believes they should. He states that, were it not for the limitations in the past
of computer speed, Monte Carlo techniques would have been used all along.
Computers have now become sufficiently fast that the time required to obtain
a typical treatment plan has shrunk to a few minutes. He argues that some
306
find the statistical jitter in Monte Carlo results troubling. He also argues that
it is hard to assess whether improvements in accuracy are clinically significant
and worth additional cost, the reason being that randomized trials in which
half of the patients are treated with less accurate methods are not feasible.
Mohans arguments therefore centre on the believe that continually reinventing
approximate dose computations and tweaking them to meet every new situation
is unacceptable and should be stopped. Broadly speaking, the Monte Carlo
techniques are applicable with the same degree of accuracy for photon treatments,
electrons and brachytherapy and, therefore, separate models are not required for
each of these modalities. The availabity of Monte Carlo calculations leads to
a dramatic reduction in the time, effort and data required for commissioning
and Monte Carlo allows the calculation of dose distributions in regions of
electron disequilibrium. Mohan argues that, because Monte Carlo techniques
are affordable and practical, it is not necessary to conduct clinical trials to
demonstrate the clinical significance of the improved dose accuracy. In summary,
provided Monte Carlo developers and users ensure that the approximations have
no significant impact on accuracy, Mohan believes that the stochastic nature of
the Monte Carlo approach is no longer an impediment to its use.
Antolak is against the proposition and believes that a fundamental difference
between Monte Carlo and analytic techniques is that the latter are deterministic,
i.e. independent calculations of the same problem will always give the same
answer, whereas Monte Carlo techniques being stochastic give different answers
each time that they are used. The error in the answer will also be proportional to
the inverse of the volume of the dose voxels and to the inverse square root of the
computational resources allocated to the problem. Antolak believes that it is not
known how much noise in the dose distributions is acceptable and also believes
that Monte Carlo algorithms usually start with a source model that requires trialand-input data adjustments to match the measured dose data. He also believes that
modelling all of the field segments for a complex dMLC fluence pattern would be
impractical under normal circumstances, although Siebers et al (2001b) and Liu et
al (2001a, b) have modelled just this situation. Antolak also believes that current
Monte Carlo treatment-planning algorithms (those being put forward for clinical
use) do introduce approximations that greatly speed up the calculations but may
affect the accuracy of the results. In summary, Antolak believes that Monte Carlo
methods should be used as an independent verification of dose delivery or to
document dose delivery but should not replace analytical methods for estimating
dose distributions in radiotherapy treatment planning. Siebers and Mohan (2003)
provide a comprehensive overview of this topic. The next paragraphs summarize
the progress made in Monte Carlo IMRT planning since 2000 and contribute to
this debate.
307
6.8.2 Determination of photon spectrum and phase space data for Monte
Carlo calculations
Monte Carlo calculations need to start from a specification of the materials
and geometry of the radiation-delivering equipment and a knowledge of the
beam spectrum. Partridge (2000) has used data taken from electronic portal
imaging devices to derive transmission measurements in materials which enable
a reconstruction of megavoltage photon spectra. The technique relied on placing
a priori constraints on the Schiff model to avoid the otherwise badly constrained
mathematical problem which would be too sensitive to noise in the source data
and non-uniqueness.
Fogg et al (2000) have developed a Monte Carlo-based phase-space source
model to predict IMRT dose distributions. Aaronson et al (2001) have also
developed a Monte Carlo-based phase space source model for accurate IMRT
dose verification. Seco et al (2003) have modelled the Elekta SL15/25 accelerator
and the Varian 2100 CD accelerator and compared the predictions of Monte Carlo
dosimetry favourably with measurement for homogeneous phantoms.
6.8.3 Comparison of Monte Carlo and pencil-beam calculations
Laub et al (2001) have compared the predictions from the KONRAD IMRT
inverse-planning system, which uses a finite-size pencil-beam (FSPB) algorithm,
with the calculations made by Monte Carlo for specific IMRT treatments of a
phantom with uniform lung inserts. It might therefore be expected that the
KONRAD planning system would produce results that would not be in agreement
with the Monte Carlo calculations. The authors, however, showed that this is not
the case and quite good agreement was obtained (figure 6.33). This is because,
first of all, the IMRT planning process tends to down-regulate the beams which
pass through the simulated lung tissue. Secondly, the lung tissue is uniform
whereas real lung tissue normally contains large volumes of air leading to electron
disequilibrium and, thirdly, selected beam directions tend to avoid the lung. So,
all in all, the conclusion is that one should still be interested in Monte Carlo
calculations for more accurate IMRT dosimetry even though in this particular
experiment they did not seem to show large differences. Laub and Bakai (2001)
have compared the dose distributions calculated with a pencil-beam algorithm
with the same dose distributions computed with the Monte Carlo code EGS4
and with measurements for an inverse plan of a thorax phantom. They found
that all three methods of measuring or computing dose agreed very closely and
that no significant overestimations of dose values inside the target volume by the
pencil-beam algorithm were found in the thorax phantom. The reason for this
as previously stated was that the dose to the low-density region lung is reduced
by the use of a non-coplanar beam arrangement and by intensity modulation.
However, Laub et al (2000a) have developed a hybrid method in which Monte
Carlo dose computations were incorporated into the inverse planning. When
308
Figure 6.33. Dose distribution of an IMRT treatment plan in the thorax region of a phantom
as calculated with a pencil-beam algorithm in KONRAD (left) and using the Monte Carlo
code (right). (From Laub et al 2001.)
tissue inhomogeneities occur, the computed fluence maps were different from
those computed with a pencil-beam dose kernel.
Wang et al (2002a) have used Monte Carlo techniques to compute
IMRT plans and noted differences between plans computed with Monte Carlo
calculations and plans computed with simpler measurement-based pencil-beam
algorithms. In general, due to computer limitations, most inverse planning uses
approximate dose-calculation algorithms. This study reported on 10 plans created
at the Memorial Sloan Kettering Cancer Center using a pencil-beam algorithm as
part of an inverse-planning technique. The fluence modulations thus created were
then converted to dose distributions once again using this pencil-beam technique.
Simultaneously, the same fluence patterns were converted to a different set of dose
distributions using a Monte Carlo technique. Five lung patients and five head-andneck patients were studied. The dose distributions obtained in the two ways were
then compared in terms of both tumour coverage and critical organ sparing. It was
found that, whilst the IMRT plans produced with the pencil-beam algorithm were
all considered clinically acceptable, dose distributions did change with the use of
Monte Carlo dose calculations. Specifically, the patient-specific images were used
to provide the electron density array from which the particle pathlengths were
evaluated and dose scoring was performed. It was found from a detailed study of
the comparisons of both isodose distributions and dose-volume histograms that,
whilst large differences were observed for the application of single fields, the
differences tended to decrease when multiple fields were in use (figure 6.34). The
largest differences were seen for IMRT plans in the lung when for one patient
there was a particularly low lung density due to emphysema. There was very
309
Figure 6.34. Comparison of the treatment plan percent isodose distributions for a
seven-field nasopharynx IMRT plan. The images on the upper row are for a standard plan
and those on the lower row for the Monte Carlo-generated plan. The PTV is the bold line
of dots. The isodose levels are labelled. (From Wang et al 2002a.)
little change in spinal cord dose for the head-and-neck plans and the total lung
doses differed slightly. In summary, the authors found that the limitations of the
pencil-beam algorithm led to varying degrees of differences depending on beam
energy, tumour location, the density of the inhomogeneity, the field sizes, the
beam directions and even the number of beams. This agrees with Laub et al
(2001) who also did not find big differences between Monte Carlo calculations
and pencil-beam calculations.
Zheng et al (2002a) have shown that the predictions of the finite-sized pencilbeam algorithm incorporated within the treatment-planning system CORVUS can
differ from the predictions of the Monte Carlo code PEREGRINE by as much as
7% when the patient has metallic implants. However, the difference is never more
than 3% even in build-up regions when comparison is made for other tissues.
Holmes et al (2000) have developed a three-stage inverse-planning process
for IMRT. In the first stage, a pencil-beam optimization scheme was used to
generate the intensity mapsthis was done with CORVUS. The second stage
310
was a forward-planning stage which used a Monte Carlo calculation to determine

dose distributions that were used in the final stage which was the beamweight
optimization stage. The whole process was very time consuming taking many
days but it increased the accuracy of dose computation to make agreement
between Monte Carlo simulation and measurement better than 1% compared with
the 34% for the pencil-beam model.
Jeraj et al (2002a) have studied the effect of using Monte Carlo-generated
beamlets in inverse planning for IMRT compared with superposition of pencilbeam calculations. Regarding the plan created using the Monte Carlo beamlets
as the accurate plan, it was shown that the use of finite-size pencil beams
can introduce a very significant (between 5% and 10%) systematic error. The
systematic error is the difference between the dose distributions computed with
the dose-algorithm and the error-free (MC) dose-calculation algorithm. An
additional error studied is the convergence error which is a direct consequence of
using the wrong beamlets. Essentially the plan converges to the optimum for the
use of the wrong beamlets rather than the error-free (MC) beamlets. The overall
conclusion of the study was that inverse-planning systems should be upgraded to
use Monte Carlo-generated beamlet doses.
Keall et al (2002b) have described how IMRT beams, computed using
superposition, the default dose calculation algorithm on a particular planning
system, were then recomputed using a Monte Carlo technique and, in 5% of the
cases, MC results in MU changes for some of the segments.
Ma et al (2000c) have shown that the FSPB algorithm in CORVUS agreed
with the predictions of a Monte Carlo computation to within 4% in IMRT
calculations where most of the tissue was homogeneous. Discrepancies up to 20%
can occur in inhomogeneous regions (figure 6.35). The Monte Carlo calculations
themselves were benchmarked to experiment to within 2%. Ma (2000) shows how
Monte Carlo is now forming the basis of inverse-planning algorithms and IMRT
dose verification.
In conclusion, there does not seem to be a consistent pattern emerging from
these studies. Under some circumastances when the dose from modulations
generated from inverse planning which uses pencil-beam distributions is
recomputed using Monte Carlo methods, the outcome shows little change.
However, the use of Monte Carlo kernels actually in the inverse planning does
seem to change the outcome.
6.8.4 MCDOSE
Some of the earliest Monte Carlo work done with a package was performed with
the EGS4/DOSXYZ code developed at Stanford. More recently, MCDOSE has
been developed. Ma et al (2002a) have described its operation in some detail. It
has been shown to be accurate to about 2% when compared with experiments.
The main feature of MCDOSE is that all the dose-modifying structures such as
jaws, physical and dynamic wedges, compensators, blocks etc are included into
(a)
311
(b)
(c)
Figure 6.35. The figure shows the dose distributions for the treatment of the vertebra
calculated by both Monte Carlo simulation and using the CORVUS system. The plan
was generated using the CORVUS system for 15 MV photon beams with nine coplanar
gantry angles (20 , 55 , 90 , 140 , 180 , 220 , 260 , 300 , 340 ). The intensity was
modulated using a Varian dynamic MLC with 80 leaves and the prescribed target dose was
18 Gy. The isodose lines shown are 17.6, 15.6, 13.7, 11.7, 9.8, 7.8, 5.9, 3.9 and 2.0 Gy
respectively in each figure. (a) shows the distribution calculated by Monte Carlo; (b) the
distribution computed by CORVUS; (c) shows the dose-volume histogram as calculated by
Monte Carlo (full curves) and CORVUS (broken curves) for the target and the spinal cord.
The Monte Carlo dose distribution shows slightly better target coverage than the CORVUS
dose distribution. (From Ma et al 2000c.)
the code. Variance-reduction techniques lead to this code operating some 30 times
faster than EGS4/DOSXYZ. It was shown how the beam parameters are tuned.
Essentially the free parameters are adjusted until the experimental measurements
match the predictions for specific beam geometries.
Ma et al (2002b) and Price et al (2003) have described how Monte Carlo
treatment planning for IMRT has been implemented at the Fox Chase Cancer
312
Center. It has been used to investigate the beam delivery accuracy for microMLCbased radiotherapy and also to study the dose-calculation accuracy for MRI-based
treatment simulation as well as other techniques including image-guided robotic
radiotherapy.
6.8.5 Speeding up Monte Carlo dose calculations
Deasy et al (2002) have developed a prototype Monte Carlo-based IMRT
treatment-planning system that uses beamlet dose distributions which are stored
in wavelet-compressed format.
Siebers et al (2000) have developed a method to speed up the Monte Carlo
dose calculation for dynamic IMRT plans. In this, incident photon weights were
calculated using the probability of photon attenuation in the moving leaf and leaf
tip and incident electron weights were modified by the probability that they do
not intercept the leaf. TG effects were also accounted for but secondary electrons
and photon scatter from the MLC was ignored. Provided the dMLC efficiency
was greater than 0.3, this approximate Monte Carlo method was not different by
more than 1% from a full Monte Carlo simulation.
Siebers et al (2001a) have commented that accurate convolution/superposition or Monte Carlo dose methods are currently considered too time
consuming to take part in iterative IMRT dose calculations. They thus proposed
a voxel-by-voxel dose-correction ratio matrix which can be applied to the pencilbeam dose results during the optimization which then aims to converge both in
terms of the inverse-plan cost function and also converge such that there is no
difference between the ratio-matrix results and the pure convolution/superposition
results.
Siebers et al (2002a) presented two methods of what they call hybrid
planning. Inverse-planning proceeded using a fast pencil-beam algorithm for
dose deposition. However, every so often, the dose matrices were recomputed
using the convolution/superposition algorithm which is known to be more
accurate. The dose matrices were corrected either multiplicatively or additively
for the differences between the outcomes of using the pencil-beam and the
convolution/superposition techniques. It was shown that this leads to a very fast
inverse-planning algorithm without sacrificing dose-calculation accuracy.
Siebers et al (2001b) compared a number of different schema for IMRT,
specifically studying issues of speed versus accuracy. Four techniques were:
(1) use of a pencil-beam algorithm throughout (fast but maybe in error);
(2) use of a convolution/superposition (CS) algorithm throughout (much [100
times] slower but said to be more accurate);
(3) recalculation by CS of the dose distribution using the fluences found in (1);
(4) as (1) but followed by a full CS optimization.
They inspected five prostate patients and concluded that
(i) techniques (1) and (2) gave quite different plans but;
313
(ii) technique 3 gave a result quite close to technique 2.

It appears the a hybrid method can generate (quicker) a solution very similar to
the full lengthy CS calculation.
(iii) Large differences in dose distributions for single fields tend to be washed out
by multiple fields.
6.8.6 Monte Carlo calculation accuracy and error
Jeraj and Keall (2000) have evaluated the effect of statistical uncertainty on
inverse treatment planning based on Monte Carlo dose calculation. They pointed
out that if inverse plans are created using Monte Carlo-generated dose kernels
(which are themselves subject to statistical noise), the dose plan will change if
the dose distribution is recalculated using noise-free data. Thus, the use of Monte
Carlo kernels actually introduces an error in the inverse planning. They call this
error the noise convergence error.
6.8.7 Application to the dMLC technique
Keall et al (2001a) have developed code to compute the dose from IMRT
delivered via the dMLC technique. The code assumed that dMLC treatments
may be modelled as a set of linked static MLC-shaped treatments. The method
accounted for intraleaf thickness variations, interleaf leakage and leaf-tip shape
in the particle transport. The MLC model modified the weight of the incident
particles by their probability of transmission which was determined from the leafsequencing file. The contribution through the open apertures, leaves and leaf tips
were calculated separately and then added. The model included the effects of
Compton-scattered photons. This was claimed to be better than direct Monte
Carlo calculation which would require the number of histories to scale in inverse
proportion to the (low) treatment efficiency. The low efficiency requires careful
calculation of leakage contributions due to to the larger number of MUs needed
for dMLC treatments. Keall et al (2001a) compared the results of the calculations
with measurements made using film for model dMLC treatments created using
the Varian 80-leaf and 120-leaf MLC operating at two energies. They showed
the agreement was better than that between collapsed-cone calculation and
measurement. The code was very effective at predicting the leakage through the
MLC leaves (figure 6.36).
Shih et al (2001) have used the EGS4/BEAM code to parametrize the phase
space of the fluence from a 6 or 18 MV beam from a Varian 2100 C accelerator
and, thence, to predict the distribution of dose from a dynamic wedge. Monte
Carlo calculations agreed with measurements to within 2% or 2 mm in the high
dose gradient.
Verhaegen and Liu (2001) and Liu et al (2001a, b) have simulated the
generation of a universal dynamic wedge for a Varian accelerator using Monte
314
Figure 6.36. This figure shows how closely measured intensity-modulated dose
distributions are to the calculations from three different planning algorithms. The isodose
and dose profile comparisons are with measurement (film) at 5 cm depth in a water
phantom: (a) Monte Carlo, (b) superposition and (c) pencil beam. The broken line
on the isodose plot indicates where the doses profile was taken. As can be seen the
dosimetric agreement between calculation and measurement was closer for the Monte
Carlo calculations than it was for the other two calculation techniques. (From Keall et
al 2001a.)
Carlo calculations. They have done this in two ways. In the first way, known
as the position probability sampling (PPS) method, a cumulative probability
distribution function was computed for the collimator position, which was then
sampled during a single simulation to create a single phase-space distribution
of fluence. In the second method, known as the static-component-simulation
method, a dynamic field was approximated by MSFs in the step-and-shoot
fashion requiring the computation and storage of a large number of phasespace fluence distributions corresponding to positions of the jaws at different
fractions of the total delivery time. It was shown that the two methods gave
comparable results and required comparable storage but that the first method
had the elegance of avoiding simulating MSFs and having to store many phasespace files. Corrections were made for the monitor chamber backscatter effect.
The overall effect is to generate the number of particle histories simulated
315
for each segment in proportion to the MUs for that segment. Calculated and
experimentally-measured dose distributions were also compared at depth dmax to
exclude scatter effects, for a sinc function in 2D delivered by either step-and-shoot
or dynamic delivery with a Varian accelerator and an 80-leaf MLC. The same was
done for a field created by CORVUS. The error on the Monte Carlo simulation
was estimated as 2.5% and the experiment was 2%. The two agreed to better than
2% for most data points which justified the approach.
Fix et al (2001a, b) have developed a Monte Carlo simulation using a
multiple-source model They modelled two step-and-shoot cases and also two
clinical treatments delivered with a dMLC. The latter were a head-and-neck and
a bronchus carcinoma. The dose profiles for all applications studied showed good
agreement between calculated and measured dose distributions. Fix et al (2002)
have commented on the fact that the delivery of IMRT with small beam segments
can lead to a change in the spectral characteristics of the beam. Using a Monte
Carlo method, they found that sweeping a small moving slit of 2.5 mm width
across the field led to an increase of the mean photon energy of up to 16% at
6 MV.
6.8.8 Monte Carlo calculations in tomotherapy
Jeraj et al (2001) have used Monte Carlo (MCNP) code to simulate the
performance of the tomotherapy prototype at the University of Wisconsin. This
was necessary because the Siemens accelerator and target used are mostly nonstandard. There is no flattening filter; instead there is a thinner target, an electron
stopper and a primary compact collimator. Good agreement was observed
between the Monte Carlo simulations and the measurements for all dosimetric
characteristics for a variety of different leaf pattern openings. The TG effect was
studied and, in addition, it was calculated that the change in the spectrum off-axis
is very minor.
Lee et al (2001a) have used the MCDOSE EGS4 user code to compute
the dose distributions arising from the NOMOS MIMiC collimator using Monte
Carlo techniques. These were compared with the CORVUS implementation of
the finite-sized pencil-beam algorithm and agreed to within 4% or 3 mm lateral
shift in isodose lines. This is a significant advance on the work done previously
by Webb and Oldham (1996).
6.8.9 Monte Carlo calculations of IMAT
Li et al (2000d, 2001c) have used Monte Carlo treatment-planning code to verify
IMAT. It was found that Monte Carlo calculations agreed with measurements in
a Rando phantom to better than 2% for absolute dose and within 3% for relative
dose distributions. However, up to 8% discrepancies were found between the
Monte Carlo calculations and the RENDERPLAN treatment-planning results,
316
largely due to the improper calculation of the effects of head scatter and tissue
inhomogeneities.
6.8.10 Other reports on Monte Carlo dosimetry
Martens et al (2001c) have shown good agreement between Monte Carlo dose
calculations incorporated into a treatment-planning system and measurements of
dose near air cavities in the mucosa.
Alber et al (2001a) have described a two-stage IMRT planning algorithm.
Inside their planning system HYPERION, a pencil-beam dose calculation is used
to create the basic segment locations of multileaves and then a Monte Carlo dose
calculation is used to make a more accurate dose calculation.
Lewis et al (2001) have developed a Monte Carlo-based system for
predicting the complete chain of events when a Varian Millennium MLC, attached
to a Varian 2100 CD Linac, is used for IMRT. The detailed model included
modelling the linear accelerator, the multileaves and incorporated CT data as well
as modelling the portal imaging detector.
Francescon et al (2001) have compared dose distributions produced by
the ADAC PINNACLE3 IMRT planning system with those produced by
EGS4/BEAM and found differences were always less than 3% for the total
treatment.
Sanchez-Doblado et al (2001) have used the BEAM/DOSXYZ code to
model the segments in an IMRT treatment and compared these with the
predictions of inverse treatment-planning systems and measurements using film.
Only when these dose maps agree can the treatment be delivered to the patient.
Hartmann et al (2001, 2002a) have developed a new inverse Monte
Carlo code for IMRT optimization and have shown the improvements in dose
calculations compared with the non-Monte Carlo plan from the Nordion HELAX
TMS planning system.
Pawlicki et al (2000a, b) have investigated the role of set-up uncertainty
in IMRT. The target volume and beam orientations were defined and then two
beamlet optimizations were done, one with and the other without the uncertainty
incorporated into the planning stage. In both cases, the beamlet weights were
optimized. Then a final dose calculation was made with the set-up uncertainty
included in each plan and the dose calculations performed by Monte Carlo
techniques. The overall conclusion was that including the set-up uncertainty in
the pre-optimized beamlet dose distributions ensured the optimal target coverage.
Li et al (2000c) have developed a water-beam imaging system for verifying
IMRT dose distributions delivered with the dMLC technique. Reconstructed dose
distributions were compared with Monte Carlo simulation results.
Papanikolaou et al (2001) have studied a large number of inhomogenous
dose-calculation algorithms applied to fluence modulations for IMRT and showed
that significant differences arose, indicating the need to always compute dose
distributions using inhomogeneity corrections.
Energy in IMRT
317
6.9 Energy in IMRT

Pirzkall et al (2000a, 2002) have studied the effect of beam energy and the
number of fields on photon-based IMRT for deep-seated targets. Ten patients
with prostate cancer were selected and IMRT plans were produced using the
CORVUS treatment-planning system. Plans were created for 6, 10 and 18 MV
photons using either four, six, nine or 11 coplanar non-opposed fields. The 12
plans produced for each patient were then compared. It was observed that the
target and sensitive structure metrics were the same for all plans, thus confirming
earlier work by Sternick et al (1997) and the conventional wisdom that IMRT
is independent of energy. However, it was found that plans with 6 MV photons
that use less than nine fields might result in an increase in dose in regions distant
from the target volume, for example near the skin surface. This is because these
regions are not included as constraints in the inverse planning and the finding is in
line with observations that such regions can occur in conventional treatment with
low-energy photons. All plans were created with five intensity levels and plans
were evaluated using standard conformality indices or metrics (figure 3.37). It
was argued that although the study was done with a particular planning system
and for a particular set of patients, the conclusions had generic applicability and
the issue of the selection of beam energy in IMRT seems to have more to do
with decreasing the cost of accelerators than improving treatment plans. Lu et al
(2000) have also demonstrated the equivalence of IMRT at 6 and 15 MV.
Papanikolaou (2001) has linked the ADAC PINNACLE treatment-planning
system for IMRT to a Varian 2100 EX linac operating in step-and-shoot mode.
Calculations were performed at 6 and 18 MV in three modes: assuming tissues
were homogeneous, 2.5 dimensional (convolution) and a full 3D scatter transport
(superposition) calculation. Differences as high as 15% in the absolute dose
were found between algorithms, particularly to OARs and it was argued that the
choice of energy appears to be of lesser importance. A low-energy beam was
recommended for practical reasons. This differs from the recommendation of
some other authors.
OBrien et al (2002) have looked again at the issue of energy in IMRT and
found that plans at 6 and 18 MV are essentially equivalent. Dong et al (2003)
have done the same and shown equivalence of plans. High energy is not required
for IMRT.
Wierzbicki and Blackmore (2003) have compared four IMRT delivery
systems, MIMiC on MDX Siemens 6 MV beam, Varian Clinac 6 MV beam
with Millennium MLC, Siemens Primus 6 MV beam with MLC and Siemens
Primus 18 MV beam with MLC. Four clinical cases were planned for each using
CORVUS. The conclusion was that the systems all performed much the same and
that high energy was not necessary.
Subramanian et al (2002) have argued for the construction of special linacs
for IMRT.
318
Figure 6.37. Plan equivalence by dose distribution. Equivalent plans in terms of the
5595% isodose lines. (Reprinted from Pirzkall et al 2002 with copyright permission
from Elsevier.)
Measuring and accounting for patient/tumour movement
319
6.10 Measuring and accounting for patient/tumour movement

Most of the development of IMRT has taken place assuming that the organs of
a patient do not move from fraction to fraction and are well represented by their
positions determined from some pre-planning 3D imaging study, be it x-ray CT,
MR or functional imaging. As the ability to conform to the target so delineated has
reached near perfection, attention is now turning to not accepting this limitation
and attempting to quantitate organ movement and to account for it in IMRT
planning and delivery. A nice soundbite could be that IMRT of the moving
patient is like completing a jigsaw on a jelly. A thought-provoking commentary
on the importance of this problem has been written by Goitein (2004) specifically
with respect to historical aspects of considering motion and also segregating the
different types of motion and their importance and methods to cope with them.
We shall see later in this section that it has been shown that if the goal is
to create a dose distribution conforming to the PTV then provided the PTV is
defined to adequately cope with all motion of the CTV within it, the IMRT dose
distribution to the PTV after all fractions have been delivered even with patient
movement is faithful to what it would have been if there had been no movement.
This is a consequence of the central limit theorem. Pursuing the analogy, it is that
none of the individual jigsaw on a jelly pieces really fit at each fraction but when
all the pieces are taken together over all fractions, the summed jigsaw fits. Pieces
from the jigsaw from one day fit with other pieces from the jigsaw from some
other day. However, this methodology considers delivering the conformal dose to
the PTV. It also makes assumptions that each fractional delivery is individually
dephased with respect to the tissue movement. To deliver a conformal dose
distribution to the CTV, other methodology is needed to either freeze movement
or track the tumour.
6.10.1 General review
Langen and Jones (2001) have put together a detailed critical review of organ
motion and its management. They have studied three types of motion. The
first is position-related organ motion which can be minimized if the patients
planning scan is performed while the patient is immobilized and in the treatment
position. The second kind of motion is interfraction organ motion, i.e. motion
that occurs when the target volume position changes from day to day. This is
mainly a problem for organs that are part of, or adjacent to, the digestive/excretory
system. This part of the review is a detailed collation of the work of all those
authors who have studied this type of motion and is classified under various
headings: gynaecological tumours, prostate (the largest group) and bladder and
rectum. Their overall conclusion for prostate is that comparisons between the
many studies are hampered by the different protocols and study end-points used.
The third type of motion is intrafraction organ motion, generally due to
respiratory and cardiac functions which disturb other organs. The review collates
320
the data for liver, diaphragm, kidneys, pancreas, lung tumours and prostate.
Specific tables indicate the studies, the number of patients, the conditions and
the average movements recorded. A number of techniques aim to minimize
intrafraction motion: e.g. synchronization of the diagnostic and treatment
procedures with breathing including voluntary breath holding, synchronized
equipment gating and forced breath holding. There are other techniques that
continually track the organ position. These are reviewed later in detail. This
very detailed paper from Langen and Jones (2001) contains a wealth of data from
some 66 clinical studies. However, it is very hard to summarize the outcome in
simple terms due to the many different conditions holding. The tables themselves
must be studied for the detail.
Several techniques have been proposed to manage interfraction motion
(Wong 2003). One is to acquire CT images just prior to treatment. The second
is the use of radio-opaque markers and realignment of the patient prior to each
fraction. The third is the use of uretheral catheters and the fourth is a kind
of adaptive process as used in the William Beaumont Hospital. These will be
reviewed later.
Image-guided CFRT and IMRT have been recognized as growth areas
(Wavelength 2001a). Image-guided radiotherapy (IGRT) attempts to image the
moving tumour online in the treatment position, during treatment or at least before
each fraction and to take account of the variation in the position of the target. Flatpanel imagers, such as that made by Elekta, and cone-beam CT are tools being
explored to take forward IGRT (see section 6.11.2).
There is much debate about whether every patient receiving 3D CFRT or
IMRT needs to receive image-based target localization prior to turning on the
beam for each fraction. In a Medical Physics Point and Counterpoint, Herman
et al (2003) debate this proposition. One view is that escalating doses or reducing
margins should not be routinely practised until evidence of the benefit of target
localization for each fraction can be produced. Another view suggests that daily
imaging introduces substantial operational costs that are only justified for some
tumour sites and should not be blindly applied to all.
6.10.2 Some observations of the effects of movement
Chuang et al (2000) have investigated the effect of uncertainties in patient
positioning and patient motion in IMRT. They found that random patient motion
and set-up error led to quite small inaccuracies but consistent set-up error such as
the chin-up motion introduced 2035% dose changes to critical structures such as
the brainstem and optic chiasm.
Verhey (2000) has compared the different delivery techniques for IMRT
including sequential tomotherapy. The techniques vary in their sensitivity to
patient motion and set-up error.
Keall et al (2002a) have proposed 4D IMRT in which respiratory gating takes
care of the movement of mobile tumours such as those in lung and breast. Keall
321
et al (2001b) have discussed synchronization of breathing motion with IMRT

delivered using the dMLC technique and shown that it is very important to cater
for breathing patterns. Such patterns can be established from analysis of gated
CT scans (see, e.g., Wahab et al 2003).
Sohn et al (2001a) have studied the effect of breathing motion on IMRT of
the breast. For each patient, an average of eight images per field were acquired
during treatment so that each patient had approximately 240 images for tangential
fields. This was done with the Portalvision imaging system. The motion of the
breast was measured and dosimetric consequences were determined.
Samuelsson et al (2003) have shown that, provided International
Commission on Radiation Units and Measurements (ICRU) margins are applied
to CTVs for head-and-neck cancers, IMRT is not particularly sensitive to
systematic set-up errors in patient position.
6.10.3 Optical imaging for movement correction
There have been many attempts to couple optical imaging of body-external
markers to patient treatment. Most, but not all, optical systems use either active
or passive infrared markers attached to the skin surface. Their motion may or
may not correlate with the movement of underlying organs and this is an issue to
consider in their use. The techniques are reviewed here.
Tome et al (2000) have used a bite-plate connected to the patients maxillary
dentition and attached to which is an array of passive infrared markers. The
movement of these markers was then detected and fed to an optically-guided
patient localization system which can correct for any variability in the patient
positioning from day to day. Tome et al (2001) have developed IMRT for
stereotactic fractionated radiotherapy based on this optically-guided system for
3D CFRT using multiple non-coplanar fields. Based on observations for four
patient cases, it was found that IMRT improved conformality indices.
Several reports have appeared on the use of the BrainLAB ExacTrac optical
navigation-and-guidance system. This is a real-time infrared tracking device for
monitoring movement. Alheit et al (2000) confirmed that the system improves
patient set-up accuracy for prostate treatment. Hagekyriakou et al (2000) used
the BrainLAB ExacTrac system to locate the prostate at successive fractions and
have found that this is accurate to 1 mm. Verellen et al (2001) have also used
a NOVALIS linac, a BrainLAB integrated miniMLC and the ExacTrac system
for stereotactic radiosurgery. The same system was also tested for irradiating
extracranial targets such as the prostate. Kim et al (2004b) showed that, for
head-and-neck patients, the ExacTrac system showed that 95% of the intrafraction
patient positions were within 1.5 mm of their planned position.
Wagman et al (2001) have evaluated the Varian real-time position-monitor
(RTM) respiration-gated system. This is a passive-marker system using an
infrared-sensitive camera to track the motion of reflective markers mounted on the
abdomen. The system can be used for both acquisition of CT images for treatment
322
planning and also, in the treatment mode, the system allows the linear accelerator
to be gated. For liver patients, it was found that motion could be decreased from
some 2.3 cm down to just 5 mm with gating.
Baroni et al (2000) have used a stereophotogrammetric technique to observe
set-up errors, breathing movement and changes in breast volume in external-beam
breast radiotherapy. The system used was the ELITE automatic optoelectronic
motion analyser. Localization errors of about 4 mm and median errors due to
patient breathing of about 8 mm were found.
Lyatskaya et al (2002) have used skin-mounted infrared reflective markers
tracked by infrared cameras with submillimetre accuracy to monitor the position
of the breast during fractions of radiotherapy. The goal was to reduce the rate of
long-term cardiac complications by irradiating the patient at deep inspiration.
Chen et al (2001a) have tracked the movement of tumours in the thorax using
fluoroscopy. They showed that there can be phase changes between the tumour
movement and the movement of surface skin markers.
Macpherson et al (2002) have designed a photogrammetric system for
measuring the intrafraction movement of the patients skin surface using two
CCD cameras. The skin surface motion was correlated to internal organ motion
captured through fluoroscopy. It was claimed that respiratory motion can be
reduced from more than 2 cm to less than 2 mm.
Ozhasoglu et al (2001) have used optical surface monitoring of chest and
abdomen, strain gauge transducers and spirometry to study the breathing patterns
of healthy subjects. They found that over an interval of 1020 min, the average
persons breathing pattern was stable enough to allow real-time detection and
correction for tumour motion. However, any successful respiratory compensation
system must detect and accommodate some breathing that is not strictly periodic
and stable.
Moore and Graham (2000) have introduced the concept of the virtual shell.
The 3D surface of the patient is derived from CT data and regarded as the correct
surface to which the patient should conform for subsequent treatment fractions.
An optoelectronic stereophotogrammetric device then measures the contour of
the patient at each treatment fraction by analysing the interference patterns of
structured light projected on to the patients skin. Then the virtual shell and the
patient surface contours are computer docked by applying small translations and
rotations to the patient position until some cost function representing the goodness
of fit is minimized, using first gross manipulations and subsequently simulated
annealing to avoid trapping in local minima. The technique was applied to 20
patients with prostate cancer who received weekly CT scans. The advantage of the
method was its non-invasiveness. It relied on earlier observations that the position
of the prostate could be correlated to the external surface of the patient (MacKay
et al 2000). The technique was verified by creating a shell for the RANDO-man
body phantom and deliberately adding surface imperfections to the shell.
323
6.10.3.1 Breast movement

George et al (2003) have quantified the effect of intrafraction motion during breast
IMRT planning and dose delivery. The effects of breathing were studied when this
was parametrized through a trace obtained using the Varian RTM system. This
trace was then adjusted to create the simulations for shallow and heavy breathing.
Using these traces, it was possible to compute the plans that corresponded to
IMRT of the breast when the patient was in four breathing conditions: no
breathing (artificial of course), shallow breathing, normal breathing and heavy
breathing. The results of this comparison showed that the PTV dose heterogeneity
increased as respiratory motion increased. The heart-and-lung dose was also
increased with respiratory motion. Conversely, the CTV dose heterogeneity
decreased as the respiratory motion increased. Due to the interplay between
respiratory motion and MLC motion during IMRT delivery, the planned and
expected doses are different. However, because of the random phase between leaf
motion and breathing, no statistically significant differences between planned and
expected doses were obtained when the planned distribution was compared with
the breathing-adjusted expected distribution averaged over 25 treatment fractions.
Korreman et al (2003) have shown that deep-inspiration breath-hold has led
to increased lung-and-heart sparing for treatment of breast cancer.
6.10.4 X-ray imaging of position
Jaffray (2003) has reviewed the field of using external x-rays to guide IMRT.
Murphy et al (2003) have treated 300 patients over seven years with the
Cyberknife at Stanford. By using the amorphous silicon x-ray detectors,
a comprehensive picture was built up of the pattern of intrafraction tumour
movement for 250 cranial, 23 spinal, nine lung and three pancreas patients.
They reported that the average variation in treatment site position was 0.45 mm
(cranium), 0.53 mm (cervical spine), 1.06 mm (breathhold lung) and 1.50 mm
pancreas. Cumulative distributions of movement were shown. However, these
histograms disguise the often observed gradual drifts in target position, drifts
which can be corrected using Accutrak (see section 6.10.8) and also the occasional
and unpredictable large movements.
6.10.4.1 Intrafraction and interfraction prostate movement measurements
Balter et al (2000) have determined the extent of prostate movement due to
breathing during radiotherapy by implanting 0.9-mm-diameter gold fiducial
markers for four patients with the markers implanted at the periphery of the
prostate and observing these with fluoroscopy. It was found, for example, that,
in the prone position, the maximum craniocaudal prostate ventilatory motion was
between 5.5 and 10.3 mm. However, this reduced to between 2.0 and 7.3 mm
in the supine position and as low as 0.5 mm with the use of a false table top
that allowed the pelvis to fall posterially. Lateral motion of the prostate was
324
negligible and anterior-posterior motion was the next most significant movement.
This potential for the prostate to move out of the high-dose volume may decrease
the efficiency of prostatic IMRT.
Kitamura et al (2002) have studied the intrafraction movement of the prostate
for 10 patients, each patient being observed five times (once a week) creating 50
datasets for analysis. A stereoscopic television system was used to observe the
internal movement of implanted gold markers within a specified region of the
prostate. Measurements were taken 30 times per second and real-time pattern
recognition was used to determine the trajectory of the gold seeds. It was observed
that there were no clear differences in movement found between the individuals.
Like Balter et al (2000), they observed that the amplitude of 3D movement was
much lower (0.12.7 mm) in the supine position than in the prone position (0.4
24 mm). The large value of 24 mm was due to bowel movement displacing
the prostate. The prostate movement is affected by the respiratory cycle and is
influenced by bowel movement in the prone position. Internal organ motion is
much less frequent in the supine position than in the prone position. The study
showed interesting 3D plots of the trajectory of the prostate gland over 2 min from
which the magnitude of the problem may be assessed (figures 6.38 and 6.39).
Nederveen et al (2002) have used a flat-panel imager to record the position
of gold markers in the prostate during a radiotherapy fraction. The markers were
small (5 mm long) but the imaging technique was capable of seeing them. Images
were taken, using just 2 MU, about every 400 ms and data were recorded for ten
patients with 251 fractions in total. Portal images were made of both anatomical
position and lateral beams to assess the motion. Individual frames in the movie
were subtracted from each other and a template-matching algorithm located the
markers. Traces were plotted of the marker positions as a function of time. Within
a time window of 2 to 3 min, the prostate moved on average 0.3 0.5 mm in the
AP direction and 0.4 0.7 mm in the cranialcaudal direction. Some individual
displacements were quite large of the order 9 mm. The authors commented
on differences between their observations and those at the Royal Marsden NHS
Foundation Trust made using MR movie loops.
Spitters-Post et al (2002) and Visser et al (2002) have implanted four fine
gold markers in the prostate and shown that, relative to bony landmarks, there
is intrafraction movement of the prostate. The markers are visible with MR, CT
and electronic portal imaging. The movement in the leftright direction had a
systematic deviation of 0.6 mm and the movement in the craniocaudal and ventrodorsal directions had a standard deviation of 0.6 and 2.4 mm, respectively. The
random variations ranged from 0.9 to 1.7 mm.
Dehnad (2001) has implanted markers in the prostate and used them to
correct IMRT treatments for random and systematic motions. Van der Heide
(2001) also implanted markers and corrected patient position using the images
of the widest component of an IMRT field. The planning was performed by the
KONRAD module within the PLATO TPS.
325
Figure 6.38. 3D trajectory of the prostate gland for a 2 min measurement period, data
recorded every 0.033 s for a patient in the supine position. X represents left/right, y
represents cranio-caudal and z represents anterior/posterior. Black dots represent the
position of the marker implanted in the apex of the prostate gland. Data for 10 patients
are separately shown. (Reprinted from Kitamura (2002) with copyright permission from
Elsevier.)
326
Figure 6.39. 3D trajectory of the prostate gland for a 2 min measurement period, data
recorded every 0.033 s for a patient in the prone position. X represents left/right, y
represents cranio-caudal and z represents anterior/posterior. Black dots represent the
position of the marker implanted in the apex of the prostate gland. Data for 10 patients
are separately shown. (Reprinted from Kitamura (2002) with copyright permission from
Elsevier.)
327
Internal target-movement monitoring almost always requires the use of

internal markers. However, Berbeco et al (2004) have produced a method for
tracking in the absence of markers.
6.10.4.2 Intrafraction and interfraction lung movement measurements
Shimizu et al (2000, 2001) have implanted 2 mm gold radio-opaque marker seeds
into a lung tumour and used a real-time tumour-imaging system to track the
movement of the lung with normal breathing. The seeds were viewed using two
diagnostic x-ray tubes and the coordinates of the markers were stored 30 times
per second. The movement was anisotropic for ungated therapy with a maximum
excursion to 16 mm. The linear accelerator was then gated to irradiate only when
the markers were within some specified range. Repeat imaging then showed the
marker movement was restricted to about 5 mm. This is the same system as used
for prostate measurements by Kitamura et al (2002) (section 6.10.4.1).
Seppenwoolde et al (2001, 2002) have used a real-time tumour-tracking
system to observe the movement of lung tumours in which gold markers were
located. When the tumours were in the upper thorax, it was found that the
amplitude of the motion could be as large as 12 mm in craniocaudal direction and
motion of both lungs and heart affected the movement of the tumour (figure 6.40).
The time-averaged tumour position was closer to the exhale position because
the tumour spends more time in the exhalation than inhalation phase. A linear
accelerator was gated to irradiate only when the markers were in a small range of
positions.
Kini et al (2001) have investigated the correlation between the movement
of internal markers and the movement of external markers. One-hundred-andfifty fluoroscopic movies were analysed for six patients for whom there was
also simultaneous monitoring of the chest wall through camera-detected infrared
markers. A simple mathematical function was found linking the external-marker
trace to the amplitude of movement but with a phase shift. This was then used aposteriori for further motion traces in which the internal motion was predicted
from the external motion and then compared with the actual measurement of
internal motion using x-rays. The variability with good coaching was about 3 mm
between prediction and actual motion. Tsunashima et al (2003) have also shown a
correlation between the movement of external optical markers and internal x-ray
markers but with a phase shift. Gulliford et al (2004) have developed an artificial
neural network (ANN) to learn and predict lung movement up to 10 s ahead of
real time in order to guide accelerator gating.
Vedam et al (2003a, b) have quantified the predictability of diaphragm
motion during respiration by using a non-invasive external marker. For five
patients and 63 fluoroscopic lung procedures, the internal position of the
diaphragm was monitored using x-ray fluoroscopy with images obtained at 10 Hz.
The position of a skin mark was simultaneously acquired using the Varian realtime position management system at 30 Hz. This was done with the patient
328
Figure 6.40. (A: left) The 3D path of the tumour for one patient during one treatment
portal: the grey dots represent the tumour position throughout the treatment, and the
black dots represent the tumour position as the tumour is detected to be inside the
real-time tumour-tracking radiotherapy system (transparent box). The asterisk represents
the planned zero position. (B: right) the average 3D curve of the tumour: the projections
on the coronal, sagittal and axial planes are drawn in thin black lines. Note that the tumour
follows a different path during inhalation than during exhalation because of hysteresis.
(Reprinted from Seppenwoolde et al 2002 with copyright permission from Elsevier.)
either freely breathing or under a variety of instructed breathing patterns. It was

shown that the internal diaphragm position correlated well with the respiration
signal and the statistical significance of the correlation was high. This correlation
was then used to predict diaphragm motion from one session to another and it
was shown that the diaphragm motion was highly predictable to within 1 mm
standard deviation independent of the method of breathing training. A limitation
of the work was that only the motion of one anatomical structure and only one
respiration signal were correlated. From this, it was deduced that the margins
needing to be added for gated radiotherapy could be significantly reduced for
lung tumours.
Blackall et al (2003) have instead created MR images at various phases
throughout the breathing cycle by gating. A high-quality reference image was
also obtained at end-exhalation and a model developed to link this vectorially
to the other images thus creating a model of lung movement. Hence, knowing
the time in the breathing cycle, the position of lung and lung tumours could be
predicted from a single MR scan. Rohlfing et al (2004) made a similar gated
MR study for liver and showed that the elastically deformable model of liver was
needed; non-rigid deformations gave the wrong location for one patient by up to
329
3 cm. Balter et al (2003a) have also developed morphological models of lung

movement to account for it in IMRT planning. Zhang et al (2003a) have used a
deformable model of the lung to create modulations tailored to lung deformation.
Lotz et al (2003) have built a similar model for characterizing bladder volume
change between scan time and treatment time.
Patients breathe asymmetrically and a good mathematical representation is

2n t
(6.6)
z(t) = z 0 b cos
for a motion in the z -direction where t = time, z 0 = mean position,

b = amplitude, is the period of motion and n parametrizes the asymmetry
(figure 6.41). The larger n becomes, the more time is spent at end expiration.
Values of b, n and can be determined from fluoroscopy. Then dose distributions
for static patients (e.g. based on a single scan at end expiration) can be convolved
with such a function to give a measure of the dose averaged over the breathing
cycle. Lujan et al (2003) have shown that it is necessary to make many
fluoroscopic measurements to adequately estimate the parameters. Also, if the
parameters predict a motion that is very different from that assumed at planning,
dose errors arise. The effect of part of the fraction time following an irregular
or different breathing pattern was also studied. This study was in the context of
CFRT of the liver at the University of Michigan.
Szanto et al (2002) have implanted gold markers in 1000 patients since 1994.
The patients were re-positioned by imaging these markers at the time of treatment
and average deviations from a reference position were usually less than 1.5 mm;
occasionally shifts as large as 1 cm have been observed.
Unkelbach and Oelfke (2003, 2004) have developed an inverse-planning
technique to take account of interfraction organ motion by the use of probability
distributions in the cost function to be minimized in inverse planning. They
specifically made use of information from a small finite number of CT scans of the
patient and addressed two intrinsic problems, that of incomplete knowledge of the
distribution of patient geometries and the sampling of this distribution by only a
finite number of fractions. They developed a technique which will create the most
suitable fluence profile corresponding to a distribution of positions for the target
geometry. They applied this technique to the classic Brahme et al (1982) problem
of a circularly symmetric uniform dose surrounded by a circularly symmetric
annulus the OAR.
6.10.5 Ultrasound measurement of position
Bamber et al (1996) have suggested that ultrasound images can be used to
measure tissue motion and indicate corrections needed for radiotherapy. At that
time of writing, the technique was not operating in 3D and not yet fast enough
for the suggested purpose. Subsequently, there have been many developments
(Meeks et al 2003).
330
Figure 6.41. Measured time course of breathing for a sample patient based on the
diaphragm motion (away from the expiration position) as observed under fluoroscopy
(triangles) compared with the model given in equation (6.6) using (a) n = 3, (b) n = 2 and
(c) n = 6. (From Lujan et al 2003.)
331
6.10.5.1 The NOMOS BAT

The main piece of commercial apparatus available is the Beam Acquisition and
Targeting (BAT) device from the NOMOS Corporation. This takes a 2D slice of
the patient and allows an organ to be contoured. The extraction of the prostate
contour from ultrasound data is not easy. Hu et al (2003) have developed an
automated technique. Then this contour is registered with that from a planning
stage and any misregistration is corrected by adjusting the position of the couch.
In principle, this corrects for interfraction motion.
The technique of using the BAT device commences by docking the probe
on the collimator so that the ultasound system knows where the isocentre is (to
which the CT data are referenced). Contours of the prostate are then displayed
in transverse and sagittal planes and moved manually to lock with the CT data,
thus giving the required three orthogonal components of movement. Anterior
posterior (AP) and lateral movements are determined from the transaxial plane
and superior/inferior (SI) movements on the sagittal images. There has been a
considerable number of studies reported using the BAT. These are now reviewed.
Lattanzi (2000) has used the BAT to measure the accuracy of position of the
prostate at different treatment fractions (figure 6.42). To establish the accuracy
of the BAT, ultrasound and CT at the same time were compared with good
correlation (better than 2 mm). Lattanzi et al 1999a, b) also showed that the
ultrasound-derived PTV correlated well with the CT-derived PTV taken at the
same time. Dong and Court (2003) also showed that the BAT measurements
closely reflect the anatomical position as determined from in-treatment-room CT.
Lattanzi et al (2000) have reported on the clinical experience of using the
BAT transabdominal ultrasound device to localize the coned-down part of the
prostate irradiated to a boosted dose (24 fractions). The technique was initiated
at the Fox Chase Cancer Center, Philadelphia in March 1999 and the study
reported on 54 patients. Prior to the commencement of the boost phase, the
patients received a new CT scan. Prostate-only fields were then determined for a
PTV with no margin whatever. Daily ultrasound localization of the prostate was
then registered to the PTV from this scan.
The averages and standard deviations of displacements found were 3.0
8.3 mm (AP), 1.86 5.7 mm (lateral) and 2.6 6.5 mm (SI). These might be
regarded as quite small, even acceptable, mean errors. However, of more note,
was the observation that 21% of AP shifts, 7% of lateral shifts and 12% of SI
shifts were greater than 10 mm. These shifts were due to the changed filling
of rectum and bladder. This emphasized that mean displacements are of little
use in characterizing actual displacements, the extrema of which can be quite
unacceptably large. By re-positioning the patient, making use of the BAT, it was
possible to have a reduction in planning treatment margins allowing safe dose
escalation. The effectiveness of IMRT would have been reduced if re-positioning
had not taken place.
332
Figure 6.42. The figure illustrates how the ultrasound probe known as the BAT is docked
to the linac machine head for the purpose of providing a 3D reference for the probe relative
to the machine isocentre. The docking plate is an easily replaceable acrylic sheet which is
fitted into the block tray. (From Lattanzi et al 2000.)
Beyer et al (2000) have also used the BAT system to study the movement
of the prostate. Twenty-four consecutive patients were monitored. Adjustments
of the isocentre were required on all patients daily and the average daily motion
was typically less than 1.5 mm. On any given day, however, more significant
adjustments were required, up to 6 mm and, indeed, 3% of readings required
adjustments of more than 10 mm, re-emphasizing the observations of Lattanzi et
al (2000). It was concluded that the BAT system is an effective means to target
IMRT in the treatment of prostate cancer.
Willoughby et al (2000) have reported on transabdominal ultrasound
localization of the prostate with the BAT system for 100 patients. It was found
that analysis of alignments performed on the first 50 patients compared quite well
to that of the last 50 indicating that the overall average shift is not changed.
Trichter and Ennis (2001) showed that the use of the BAT system detected
significant prostate motion unrelated to the position of bony anatomy that was
detected using EPID images. It was concluded that the BAT system provided
improved positional information for IMRT delivery.
Falco et al (2001) have used the BAT system to determine that, over a course
of fractionated treatment to patients with prostate cancer, the prostate can shift by
as much as 1 cm from its reference position in any direction. Daily displacements
were then incorporated into the treatment-planning process to assess changes in
the target dose coverage.
333
Chandra et al (2001, 2003) have reported their clinical experience of

ultrasound-based daily prostate localization at the MD Anderson Cancer Center
in Houston. For 147 consecutive patients, 3509 BAT alignment procedures were
made and reported. Ultrasound images were scored depending on their image
quality. The ultrasound image quality was judged to be poor or unacceptable
in 5.1% of the cases. Of the remaining BAT images with high-quality imaging
scores, alignments were unacceptable (> 5 mm error as judged by the reviewing
physician) in 3% of cases. The mean shift in each direction of the prostate
averaged over all patients was quite small, 0.50.7 mm. However, the distribution
of shifts was a near random Gaussian in all three major axes and the standard
deviation of the spreads was found to be 2.8 mm in the right/left direction,
4.9 mm in the AP direction and 4.4 mm in the SI direction. The percentage of
BAT procedures in which the shift was > 5 mm was 28.6% in AP, 23% in SI
and 9% in RL directions. The BAT ultrasound system corrects for both set-up
uncertainty and internal organ motion, thus compensating for inadequate initial
set-up using skin marks. The use of the BAT procedure added an extra 5 min
to each treatment slot. By plotting regression plots, it was possible to separate
the positioning errors into random and systematic errors. Interestingly, the couch
shifts did not vary significantly as the study progressed, showing that there were
no significant changes over time. It was hypothisized that, after extensive use
of the BAT alignment procedure, the therapist may have paid less attention to
the initial set-up using skin marks and relied more directly on the BAT for final
position adjustment thus leading to large recorded shifts. It was concluded that,
conversely, intrafraction prostate movement was not of grave concern.
Serago et al (2002) have also investigated the use of transabdominal
ultrasound to localize the prostate and its immediately surrounding anatomy and
guide the positioning of patients for the treatment of prostate cancer. They used
the BAT system and evaluated the same-day repeat positioning by the same
ultrasound operator and also the variation in the measurements made by two
different operators. Self-verification tests of ultrasound positioning indicated a
shift which was < 3 mm in more than 95% of cases. Interoperator tests indicated
shifts of < 3 mm in 8090% of the cases. A comparison was also made with
conventional localization based on CT and skin markers and the mean difference
in patient positioning between conventional and ultrasound localization for lateral
shifts was 0.3 mm, vertical shifts 1.3 mm and longitudinal shifts 1 mm. However,
on a single day, the differences could be > 10 mm for 1.5% of lateral shifts,
7% of longitudinal shifts and 7% of vertical shifts. This rather emphasizes that
stating the mean variations is not as adequate as stating the range and maximum
variations. It was also found that the pressure of the ultrasound probe displaced
the prostate in seven out of 16 patients by an average distance of 3.1 mm.
However, interestingly 56% of the patients showed no displacement. The quality
of this work is very dependent on the quality of the ultrasound images.
Morr et al (2000, 2002) have used the BAT system for prostate localization
and alignment with IMRT fields at each treatment fraction. Twenty-three patients
334
were studied, 19 of whom were successfully imaged supine with a full bladder.
The BAT system was then used to create ultrasound images of the prostate and
these were registered to contours previously determined from CT data. After
training, this could be done in about 6 min. The average rightleft, updown
and inout adjustments were 2.6, 4.7 and 4.2 mm, respectively. Some positional
adjustments larger than 10 mm occasionally occurred. On average, the time
required for verification and position changes was less than 10 min with an
average of 5.56 min. Positional adjustments > 10 mm were very rare and related
mainly to a misidentification of the target structures on the ultrasound image or
patient movement.
Heon et al (2002) have used the BAT system to monitor the positional
information for 22 prostate cancer patients and 504 ultrasound scans were
obtained and analysed. The prostate displacement data showed that the prostate
can shift from its intended treatment-planning position by median values of 3 3,
5 4 and 5 3 mm, in the axial, sagittal and coronal axes, respectively.
Little et al (2003) have used the BAT system to show the mean and range
of prostate motions and related this to the margins required on the CTV to ensure
tumour coverage for all fractions. The prostate organ motion shift was defined as
the difference between the BAT shift and the set-up error shift defined on portal
films taken at the same time as BAT measurements. Significant numbers of patient
treatment fractions showed the prostate located outside the PTV and so requiring
post-BAT intervention (repositioning) to restore irradiation fidelity.
Van den Heuvel et al (2003) showed that the use of the BAT system does
not improve on positioning. They implanted five gold seeds in the prostate and
imaged them each fraction for 15 patients. BAT repositioning was performed on
alternate fractions only, allowing assessment of the positioning accuracy with and
without ultrasound repositioning. Whilst there were differences, it was concluded
that the residual errors did not greatly change the positioning accuracy. The only
significant improvement was in the craniocaudal direction. Since this study is
negative to the promise of the BAT system a detailed study of its conclusions is
recommended. Langen et al (2004) also found that the BAT did not reduce the
errors in the SI and lateral directions but only in the AP direction. They found
that the BAT did not agree with x-ray images of implanted markers.
All these studies specifically concern interfraction prostate motion. It
probably makes no sense to provide here averages for interfraction prostate
movement detected with the BAT because the experiences are varied. Possibly
the recorded values reflect local practices. There certainly appears to be some
effect of learning, not all of it beneficial. The reader can make up their own mind
from the statistical data given here for the several centres that have used the BAT
system.
Huang et al (2002a) have used the BAT system to quantify the intrafraction
motion of the prostate (quote marks because this was not a continuous
measurement during the treatment). This was done by taking ultrasound
snapshots before and after each fraction for a total of 400 BAT procedures. It was
335
found that the mean intrafraction movement was negligible, i.e. 0.01 0.4 mm
in the left direction, 0.2 1.3 mm in the anterior direction and 0.1 1.0 mm in
the superior direction. There were some individual movements much larger than
this. Moreover, the movements did not correlate with interfraction movement. It
was commented that others, using fluoroscopy over a period of time, had seen
larger movements and this may be a weakness of the before-and-after snapshot
ultrasound study.
Mah et al (2001) have deduced from MRI cine video loops of patients
with prostate cancer that no significant respiratory motion could be deduced in
terms of prostate displacement. The prostate appears to dance as a result of
peristaltic motion of the rectum but only with an amplitude of less than 2 mm
which is usually of the order of the set-up error. These findings were confirmed by
Huang et al (2001b) who used the BAT system to measure intrafraction prostate
motion during 100 IMRT treatments for prostate cancer. They noted that the
movement was always less than 5 mm. Kitamura et al (2001) come to the reverse
conclusion that intrafractional movement of the prostate, measured using a realtime tumour-tracking radiation therapy system imaging implanted gold markers,
actually shows that intrafraction movement can be as large as 8 mm.
6.10.5.2 Other ultrasound systems developed
Bouchet et al (2000) have developed a 3D ultrasound technique for high-precision
guidance of radiation therapy. A Voluson 530D 3D imaging system was used to
create ultrasound images and the ultrasound probe position was tracked using
a CCD camera which is focused on four infrared light-emitting diodes attached
to the ultrasound probe. This then allows the ultrasound images to be precisely
locked onto x-ray, CT or MR scans and displacements at the time of treatment
can be corrected using this technique. The claimed accuracy is better than 1 mm
in AP, lateral and axial directions.
Sawada et al (2002, 2004) have used real-time ultrasound measurement
during treatment to correlate images with equivalent ultrasound images created
at the planning stage. A CT scanner and linac are in the same room and an
ultrasound reference image is recorded at the time of CT scanning. Then the
patient is rotated by 180 to the linac. The ultrasound probe is on a firmly
anchored robotic arm and then proceeds to record images of the patient on the
linac (figure 6.43). The images are correlated with the reference image and a
trigger pulse to the linear accelerator is generated when the image correlation
index exceeds a predetermined threshold level. This is a form of gating based on
the use of ultrasound.
Artignan et al (2002b, 2004) have studied the effect of the pressure of
an ultrasound probe making measurements of the position of the prostate but
unwantedly disturbing the prostate during the measurement. When the probe
moved from 0 to 35 mm, the prostate moved from 0 to 10 mm so the abdominal
pressure actually creates prostate displacement (see also Serago et al (2002).
336
Figure 6.43. Diagram of a system comprising a CT scanner, a 3D ultrasound device, an

articulated digital localizer and a linac. The CT scanner and the linac are located in the
same treatment room so that a tumour position detected by the CT scanner coincides with
the isocentre of the linac when the couch is rotated around the couch support by 180 . The
probe of the ultrasound device is attached to the end of the localizer arm. The system allows
accurate repositioning of the patient at the treatment time and the recording of ultrasound
images to correlate with those at the CT scanner for gating treatment. (From Sawada et al
2004.)
Ultrasound images were made of ten healthy volunteers using a transabdominal

ultrasound probe with a precise method of indexing the probe relative to the skin
surface to quantify pressure. The probe was set to point inferiorly avoiding too
much shadow from pubic symphysis and to obtain a clear view of the prostate.
The position of the prostate corresponding to light pressure was determined as
reference, even though it was hard to view the prostate (the reason more pressure
is needed). Then the distance of the probe from the prostate was shortened in
carefully controlled 5 mm intervals. The volunteers did not all behave the same
way but in all volunteers the greatest prostate movement was in the posterior
direction. On average, there was 3 mm movement for every 1 cm of applied
pressure. The prostate motion was less than 5 mm in 100% of the volunteers
after 1 cm of pressure, in 80% after 1.5 cm, in 40% after 2 cm, in 10% after
2.5 cm and in 0% after 3 cm of pressure. No correlations were found with
bladder volume or prostate-to-probe distance. There was some motion inferiorly
but random motion leftright.
Conversely, the work of McNeeley et al (2003) disputes this. They made
MR images of the prostate with and without simulation of the pressure of an
ultrasound probe and showed the prostate moved only about 1 mm on application
of the probe.
337
6.10.6 Magnetic monitoring of position

Overcoming the problem of the moving tumour is the next challenge in IMRT.
Seiler et al (2000) and Muench et al (2001) have proposed a novel technique
to track the real time position of the tumour with respect to the radiation beam
(figure 6.44). It relies on the principle of coupling a source of magnetic field to
a position-sensitive sensor. This sensor could be attached to the surface of the
patient if the tumour were superficial or, alternatively, embedded surgically to be
close to a deep-seated tumour. Organ motion generated by breathing, circulation,
peristalsis etc can then be tracked. Experiment showed an acuracy of 12 mm
or 0.51 of rotation. A phantom which was moving was then irradiated using
this technique to gate a proton irradiation and the resulting dose distribution was
very similar to that found for irradiating a stationary phantom (but quite different
from that irradiating the movement-untracked phantom). Balter (2003) has also
described a technique for monitoring the position of an implanted transponder
using electromagnetic fields with reported accuracy to 2 mm at 10 Hz frame rate.
Balter et al (2003b) have also made magnetic measurements of position using the
apparatus from Calypso Medical technologies of Seattle and shown that absolute
errors were of the order 1 mm or less.
6.10.7 Gating
One of the potential dangers with IMRT is that dose distributions which have been
tightly conformed to a target may be inappropriately placed if the target is moving.
The use of target margins is one traditional way to accommodate this but is not
a way to effectively overcome the problem. We should remember that margins
are an artificial construct invented by physicists to solve specific problems. In an
ideal world we should find a way to do without them. There is a growing interest
in the use of gated IMRT. The technique works like this. The breathing motion of
the patient is monitored and converted into a binary on/off signal. For example,
this signal could be set at a particular state of inspiration or expiration. The signal
so generated then gates the radiation either on or off with the net result that the
target is in almost the same place for each of the gated sequences of radiation.
This, unfortunately, leads to a very small duty cycle for radiotherapy.
6.10.7.1 Gating based on optical measurements of surface markers
Solberg et al (2000) have investigated the feasibility of gated IMRT. Respiration
was monitored in real time using cameras which track infrared-reflecting markers
placed on the patients surface and gating occurs for a period around exhalation.
Radiation was delivered using a Novalis accelerator and integrated microMLC.
The advantage of this system is that the beam can be started and stopped within
0.01 MUs.
Kini et al (2000) have described a gating apparatus (a passive marker
video-tracking motion-monitoring system) to improve the treatment of thoracic
338
Figure 6.44. A schematic diagram showing how the tumour-tracking technique using
miniaturized sensors with the help of magnetic fields takes place. The patient carries
an implanted sensor which moves with the tumour. A magnetic field is generated via a
six-coil tetrahedron-shaped assembly and the field sensor is a miniaturized induction coil.
The alternating magnetic field created by the field generator induces an alternating voltage
in the sensor which is measured by the connected data acquisition electronics. (From Seiler
et al 2000.)
malignancies such as lung and oesophageal carcinoma. Phantom studies have

shown that, using respiratory gating, the PTV could be decreased substantially.
Ramsey et al (2000) have shown that respiratory gating can improve the
treatment for lung cancer patients. A prototype respiratory-gating system was
used which correlates the signal from a video-based respiration-monitoring
system with the internal location of the target volume. Upper and lower gating
thresholds were determined using free breathing by the patients and then those
patients whose observed target motion under fluoroscopy was greater than 0.5 cm
were assigned to gated treatment paths. It was found that the average beam-on
time for respiratory-gated treatment increased by a factor of 2.8.
Vedam et al (2001) have determined parameters for respiration gated
radiotherapy. Their method determines the amplitude of motion and accounts
for any difference in phase between the internal tumour motion and the external
marker motion. An infrared camera tracks movements of external markers and
a fluoroscopic system creates images of internal markers (figure 6.45). The
movements are then tracked to determine amplitude and phase differences. The
339
Figure 6.45. A volunteer in a simulation position showing the infrared reflective markers
placed mid-way between the umbilicus and the xyphoid process. The camera with an
infrared illuminator surrounding the lens is inset. (Note that for patients the marker block
is attached directly to the skin rather than to the clothing.) (From Vedam et al 2001.)
choice of gating techniques is then whether to gate at inhale or exhale and then the
second decision is whether to track using amplitude or phase. The third decision
is to determine the cost benefit of how long one wishes to extend the delivery
time for the sake of decreased tumour motion. All gating techniques assume
reproducibility from breathing cycle to breathing cycle. It was observed that
gating during exhalation was more reproducible than gating during inhalation.
Ford et al (2002a) have evaluated a respiration gating system using film
and electronic portal imaging. The gating system measures respiration from the
position of a reflected marker on the patients chest. Simultaneously fluoroscopic
movies were recorded. It was found that the use of the gating system reduced the
localization of dose by about 5 mm.
6.10.7.2 Gating based on x-ray fluoroscopic measurements
Shirato et al (2000a, b, 2004) have described a technique for real-time tumour
tracking and gating of a linear accelerator. The technique relies on four sets
of diagnostic x-ray television systems which offer an unobstructed view of the
patient (figures 6.46 and 6.47). The system measures the location of a 2 mm
gold marker in the body 30 times a second. The marker is inserted in or near
the tumour using image-guided implantation. This system allows the tracking of
markers placed in the prostate, in lung tumours and in liver tumours. The linear
340
Figure 6.46. The configuration of a real-time tumour-tracking system synchronized to a

linear accelerator: (1) linear accelerator; (2) internal marker in the patients tumour; (3)
the patient couch; (4) high-voltage power units, x-ray tubes, image intensifiers. (Reprinted
from Shirato et al (2000a) with copyright permission from Elsevier.)
accelerator is gated so that the treatment beam is never on at the same time as
the diagnostic x-ray units are pulsed and the gating of the linac is performed by
means of a grid on the electron gun. The system allows the determination of the
3D coordinates of the tumour marker and the linac was energized only during
that period in which the detected location of the tumour marker was within the
accepted volume as defined by the allowed displacement value.
A phantom with a moving marker was used to establish that the accuracy
of determining the displacement was better than 1.5 mm provided the tracked
marker moved at constant velocity. From measurements made with patients, it
was determined that the range of the coordinates of the tumour marker during
irradiation was 2.55.3 mm whereas it would have been 9.638.4 mm without
tracking. It was established that the dose due to the diagnostic x-ray monitoring
was less than 1% of the target dose and was thus an acceptable additional
radiation burden. Shirato et al (2000a) claimed that their technique has some
advantages over others. For example, active breathing control requires a measure
of cooperation of the patient; optical trackers can only look at surface markers
which may or may not correlate with the position of deep-seated implanted
markers. Magnetic tracking of a single marker coil requires somewhat invasive
wiring. The main limitation of their technique was the potential for migration
of the implanted marker. Another issue is that responding to image data with
actions affecting the radiation has inherent delays which have been characterized
341
Figure 6.47. The motion-gated linear accelerator system and fluoroscopic real-time
tumour-tracking system. Three of the four fluoroscopic systems are shown. (Reprinted
from Shirato et al (2000b) with copyright permission from Elsevier).
by Sharp et al (2003). Provided the imaging is at greater than 10 Hz, there was
no advantage to using prediction techniques. At slower frame rates, computer
prediction was required.
6.10.7.3 Imaging and therapy gated by respiration monitor

Mageras (2001) and Mageras and Yorke (2004) have used respiratory gating (RG)
for both imaging and radiation therapy. For CT and electronic portal imaging,
acquisition was triggered by the RG system, whereas, for PET and fluoroscopy,
image data were tagged or time stamped with phase from the respiration
waveform and then selectively viewed according to phase. Mageras (2002)
created respiration-correlated spiral CT images to achieve 4D imaging. The x-rayon signal from the scanner and the respiration waveform from a position-sensitive
respiration monitor placed on the patient were recorded synchronously during
acquisition and used to correlate the CT slice with the respiration phase. Then,
by selecting reconstructed slices that occured at the same phase in the respiration
waveform, 3D images were created at typically 810 different phases and these
give information on the anatomical variation with phase of the breathing cycle.
These systems allowed gated IMRT delivered with either the dynamic or step-andshoot multileaf collimation. Fluoroscopy showed that external monitor movement
correlated well with that of the diaphragm in most patients.
342
Wong (2002) has highlighted two techniques for compensating for breathing
motion during the delivery of IMRT. The first is respiratory-gated radiation
treatment (RGRT). Typically, the relatively quiescent segment near end exhalation
is chosen with a duty cycle of about 25%. The disadvantage of this technique is
this low duty cycle and the need to still apply margins for the movement during
the radiation-on period. The second technique is active breathing control (see
section 6.10.11) which, in principle, should result in a smaller margin. However,
at present, it has not been determined how much margin reduction can be safely
achieved and this remains a research priority.
Jiang and Doppke (2001b) have studied the effects of respiratory motion
on the treatment of breast cancer with tangential fields. A spirometer was used
during CT scanning so that CT scans could be reconstructed at three instances of
respiration status. The first was during normal breathing and this was used for
treatment planning. The other two sets of data were acquired by breath-holding
at inspiration and expiration. When the tangential fields were shifted relative
to the CT data sets to simulate the effects of this motion, it was found that the
results suggested that breast motion during normal breathing may not be a severe
problem in tangential treatment.
Giraud et al (2000) have evaluated intrathoracic organ motion during the
physiological breathing cycle in order to optimize the 3D CFRT of lung tumours.
The patient breaths through a mouthpiece connected to a spirometer. Significant
differences were noticed in CT scans taken at different parts of the breathing
cycle and it was suggested that a spirometer-gated irradiation would be able to
overcome the problems posed by intrathoracic organ motion during treatment.
The signal from a spirometer is directly correlated with respiratory motion
and is, therefore, ideal for target respiratory motion tracking. However, it is
susceptible to signal drift which deters its application in radiotherapy. Zhang et al
(2003b) have developed calibration techniques to enable the spirometer to provide
a long-term drift-free breathing signal and it was concluded that spirometer-based
monitoring is most suitable for deep-inspiration breath hold and less important
for free-breathing gating techniques.
Van Herk et al (2002a) have created respiration-correlated CT scans for lung
cancer patients. A small thermometer is placed under the nose and the temperature
difference between inhaled and exhaled air was used to detect the breathing phase.
The thermometer signal was correlated with the CT scanner by simultaneously
digitizing the x-ray-on signal. This allows a set of 200300 raw CT scans acquired
in 35 min to then be grouped, with each group containing just those 40 slices
with irregular slice spacing corresponding to a selected breathing phase. The
volumetric images are then interpolated and are useful for showing the resulting
tumour motion curves demonstrating the precise 3D path of the tumour.
Van Herk (2003) has described two methods of image registration for
assisting radiotherapy planning. The first is a pixel-by-pixel registration in which
the comparison is based on a so-called cost function describing the goodness of
fit between the two scans. The second is a volume registration algorithm which
343
is based on registering previously defined contours, for examples bony anatomy

(chamfer matching).
6.10.7.4 Measurements using oscillating phantoms

Kubo and Wang (2000) have investigated whether IMRT dose distributions are
suitably maintained when the additional constraint of linear accelerator gating is
brought into play. Square pulses were used to mimic breathing at regular intervals
and also pulses which were generated from the spirometric curve characterizing a
patients real breathing. Experiments were carried out (figure 6.48) to irradiate a
film using either an enhanced dynamic wedge or an intensity-modulated fluence
created with a dMLC technique. A reference measurement was made with
the film stationary with respect to the moving leaves and jaws. Secondly, an
irradiation was made with the film oscillating as a result of breathing motion
with respect to the same irradiation conditions and, thirdly, a measurement was
made with the film moving under breathing conditions but with the radiation
gated using the signals previously described. The measurements were done at
different accelerator pulse-repetition rates80, 240 and 320 MU min1 and for
two energies6 and 18 MV. Motions of the film were either in the direction of
the leaf motion or at right angles to this. Kubo and Wang (2000) showed that
the gating results (the third set) were almost identical to those with no gating
and no motion, whereas there were large differences between the results from this
third set of irradiation conditions and the dosimetry measured with no gating. The
results were largely independent of beam energy, the output rate of the accelerator
or the direction of movement of the target. It was therefore concluded that linearaccelerator-gated operations maintain the enhanced dynamic wedge and IMRT
dose delivery.
Hugo et al (2001) have compared dose distributions for IMRT delivered to
a stationary target with IMRT delivered to film in an oscillating target with the
oscillating motion removed by gating the accelerator using respiratory triggering
based on infrared photogrammetry. No significant dose differences were found
between stationary non-gated films and oscillated gating films. Hugo et al (2002,
2003a, b) showed that the addition of gating reduced the maximum distance to
agreement and dose difference from 4.6 to 0.61 mm and from 15.1% to 4.45%
respectively.
Hugo et al (2002) have evaluated the use of a BrainLAB ExacTrac
commercial patient positioning system for gating IMRT (figure 6.49). An
experimental arrangement was constructed whereby a phantom could be
oscillated to simulate the breathing function that was derived from the motion
of a liver. IMRT was delivered either with single or multiple fields, both with the
MSF and dMLC techniques, to the stationary and the moving phantom. IMRT of
the moving phantom was gated with a variety of window sizes. It was found that
with the smallest window size the error as measured by the factor was lowest
344
Figure 6.48. The experimental set-up for film motion using a respirator: (A) 6-cm-thick
solid water, (B) 1.5-mm-thick spacer, (C) film, (D) lead blocks to restrict the film-holding
arm from drifting sideways, (E) reflective marker in its holder, (F) diaphragm which moves
the film-holding arm in the direction indicated by a double arrow and (G) air inlet from an
respirator. (From Kubo and Wang 2000.)
Figure 6.49. The experimental set-up for single field studies involving motion of the
phantom with respect to the beam delivery. Two slab phantoms (RSD dry water) were
placed on the positioning stage to provide build-up and localization. Film was placed
between these phantoms and markers were attached to the build-up material. The phantom
could then be oscillated with respect to the beam. (From Hugo et al 2002.)
345
and although the duty cycle was quite small (7%) a 1 mm window placed centrally
on the exhalation phase was recommended.
Goddu et al (2002) have used an oscillating phantom to show the difference
between dose distributions with and without breathing delivered with IMRT using
a Varian accelerator.
Suh et al (2003, 2004) transferred the patient movement information to the
MLC fields for IMRT so that the leaf patterns follow the tumour breathing. Yi
(2004) and Suh et al (2004) have developed a method to compel a particular
breathing pattern using a ventilator. Then it was shown that if an established
breathing pattern could be applied to MLC motion, the beams effectively tracked
the tumour and the irradiation was as accurate as applying unmoving beams to
an unmoving target. This was shown to be effective for both simple unmodulated
fields and also for step-and-shoot IMRT components. The direction of MLC leaf
movement was arranged to coincide with the direction of the movement of the
phantom. It was also assumed that the breathing motion was regular. For patient
implementation, the breathing motion was compelled to a fixed rhythm set by the
patient themself and also corrections were made for superiorinferior breathing
motions by arranging for the MLC leaves to track in this orientation (see also
Keall et al 2003).
Some other experiments with oscillating phantoms are presented in
section 6.10.12.3.
6.10.7.5 Evidence against the need for gating
Zygmanski et al (2001a) took fluence distributions from the HELIOS inverseplanning system and converted them to effective incident fluence taking into
account the expected motion of tissues due to patient respiration. These effective
incident fluences were replanned onto the patient to demonstrate the effects of
IMRT treatment with and without respiratory gating. It was found, surprisingly,
that respiratory gating did not significantly improve the treatment. Zygmanski et
al (2001c) studied the problem of IMRT delivered with a dLMC system in the
presence of organ motion. They observed that gating effectively eliminates organ
motion but extensively prolongs the delivery time, whereas tumour tracking, an
alternative solution, requires real-time organ imaging. The dose error was studied
for a periodically moving target resulting from IMRT delivered without gating or
tumour tracking. IMRT plans for the pancreas and lung were made with HELIOS
and transferred to a phantom which was firstly stationary and secondly resting on
top of a motorized table executing periodic motion simulating organ motion with
an amplitude of 1 cm and a periodicity of between 515 s. Although individual
IMRT fields showed dose differences between 10% and 30%, composite fivefield plans exhibited dose differences of between 5% and 15% indicating the
possibility that dMLC IMRT treatments can be used with caution without gating
under conditions of normal breathing. Mechalakos et al (2004) also observed
that, with suitable margins, the majority of lung tumours are covered during
346
conventional lung radiotherapy despite movement although it was conceded that

large respiration-induced motion requires further strategies for correction.
6.10.8 Robotic feedback
The development of CFRT and IMRT has provided the specific challenge of the
problem of irradiating the moving tumour. There is a considerable history of
research into techniques to monitor tumour movement. Many of the ways to
improve therapy concentrate on either gating radiation therapy to specific periods
in which the tumour is known to be in the same position or treating the envelope
of tumour movement. The latter approach involves the use of margins which
undoubtedly over-irradiates normal tissue.
A dramatic new development is robotic radiation therapy in which, instead
of the use of a conventional linear accelerator and couch, the in-line high-fluence
linear accelerator is carried on a industrial robotic arm and can point with six
degrees of freedom towards the tumour (see section 4.1).
It has been pointed out (Webb 1999, 2000e) that this particular form of
radiation therapy is particularly amenable to coping with the problem of tumour
movement. If the movements can be tracked, they can be fed back to the
robotic arm to effectively track the moving tumour. Schweikard et al (2000) have
performed some pioneering experiments to correlate the position of a tumour as
determined by stereoscopic x-rays with the position of the external skin surface
as determined by infrared sensors. The rationale behind this work is to recognize
that, ideally, one would like to monitor the internal position of tumour markers
using x-rays throughout the irradiation but that this is not possible for reasons of
unwanted radiation dose. Instead the 3D position of tumour markers is correlated
with the 3D position of external surface markers to generate a mathematical
relationship between the two. Then, throughout the treatment, the position of
the external markers is monitored and this relationship used to re-compute highframe-rate locations of the internal markers and feed this movement data back to
the robotic arm for compensation. The technique is even cleverer than this. It
can continuously update the relationship between internal and external markers to
take account of changes which may gradually take place during the treatment. It
can also distinguish between changes in tumour position that are due to breathing
and internal movements from those which are due to overall global shifts in the
patients position due to coughing, twitching or other involuntary movements.
This method of robotic radiation therapy has been developed for the Accuray
treatment-delivery system at Stanford, CA (Accutrak).
Hough and Jones (2002) have developed a robotically controlled treatment
couch and chair for patient positioning for proton therapy. They indicate that this
would also be a suitable alternative to the current couch technology for photon
technology.
Suh et al (2002) have alternatively observed that there is a strong correlation
between the movement of the skin surface and the target motion for tumours in
347
the lung. Therefore, they consider it is possible to predict the exact target location
from the skin motion which will be useful to aid gated radiation therapy. Ten
patients were imaged lying down on the simulator with radio-opaque markers on
their skin and the fluoroscopic images were recorded simultaneously.
Yan et al (2004) correlated the motion of infrared passive skin markers
detected with the BrainLAB ExacTrac system and fluoroscopically-detected
motion of internal markers. Provided the phase shift was found, there was a good
cross-covariance.
6.10.9 Held-breath self-gating
Another technique to control tumour movement in the upper thorax is to irradiate
only at one specific phase of the breathing cycle with the patient themselves
controlling this. Ideally, this also requires 3D imaging of the patient at this same
breathing phase. Forster et al (2003) have shown considerable differences in
lung position determined from free-breathing CT scans and from gated CT scans.
Rietzel et al (2003) have also shown that gated CT (4DCT) leads to clearer
definition of specific breathing phases.
Mah et al (2000) have evaluated technical aspects of the deep-inspiration
breath-hold technique in the treatment of thoracic cancer. In this technique, the
patient was verbally coached through a modified slow vital-capacity manoeuvre
and brought to a reproducible deep inspiration breath-hold level. The goal was
to immobilize the tumour and also to expand normal lung out of the high-dose
region. It was inferred from data on the first seven patients that the displacement
of the centroid of GTV from its position in the planning scan over some 350
breath-holds was only 0.02 0.14 cm. In this particular technique, the patient
held their breath for up to 10 s during a high-dose-rate irradiation. The manoeuvre
for deep-inspiration breath-hold as practised at Memorial Sloan Kettering Cancer
Center cannot, however, be performed by 60% of patients with lung cancer
(Mageras 2004).
Kim et al (2001) have evaluated the feasibility of a held-breath self-gating
technique in radiotherapy of lung cancer. Sixteen consecutive patients with nonsmall-cell lung cancer were accrued for the study and the patients were asked to
hold their breath at four points in the breathing cycle after a standardized training.
These were: maximum and end tidal, inspiration and expiration. While under
fluoroscopic visualization, it was found that maximal inspiration and expiration
tended to provide the best positional reliability and the standard deviation of
diaphragmatic position ranged from 0.13 to 2.57 mm with an average of 0.97 mm.
The day-to-day variation of diaphragmatic position was less than 5 mm and
the held-breath self-gating technique resulted in a reduction of diaphragmatic
movement by an average of 11.9 mm when compared to that seen with the tidal
breathing.
Barnes et al (2001) have studied 10 patients undergoing radiotherapy for
stage 1-3B NSCLC with and without deep-inspiration held breath. This technique
348
of immobilizing the tumour is self-gating with respect to the radiation. With

appropriate margin selection, the use of deep-inspiration breath-hold reduced the
percent of lung receiving more than 20 Gy from 12.8% to 8.8%. Patients showed
a remarkable variability in the duration of breath-hold (1952s). Della Biancia et
al (2003) have shown that IMRT at end inspiration generated lower normal lung
dose than IMRT at end expiration.
Onishi et al (2001) have taken a similar approach to controlling respiratory
motion. They asked the patient to hold their breath at deep inspiration and
provided the patient themselves with a switch which directly controlled the output
of the linac. This was argued as having very high time-efficiency because the
patient only operated the switch for the period in which they could hold their
breath.
Shimoga et al (2001) have developed SmartGate. This is a standard
stethoscope placed on the neck with a radio transmitter coupled to a receiver
outside the treatment room. The equipment is capable of distinguishing between
different parts of the breathing cycle and, in particular, permits gating-out times
in which the patient is breathing abnormally or when the patients breathing is
interrupted by nasal urges such as coughing and sneezing. The system uses an
expert rule base and human-like fuzzy-decision capability.
Ozhasoglu et al (2003) have shown using the Cyberknifes imaging system
that natural breath hold can yield lung tumour positions reproducible to within
2 mm. Ma (2003) has also commented on the positioning accuracy of the
Cyberknife.
6.10.10 Intervention for immobilization
Lofroth et al (2000) have used a urethral catheter to localize the prostate with
high precision during treatment. The technique has been used with more than 250
patients and allows margins to be reduced without external technical support.
Soubra et al (2001a, b) treat the prostate with the patient prone, lying in
a vacuum bean bag, inside a specially designed wooden cradle. The prostate
movement was also reduced by inserting a balloon into the rectum to a specified
location, inflated to 50 cm3 of air and patients were treated with the bladder full.
It is claimed that this leads to a reproducibility of patient position to better than
5 mm.
6.10.11 Active breathing control
The first active breathing control (ABC) device built at the William Beaumont
Hospital, Royal Oak, was based on a modified Siemens servo ventilator 900C
(Wong 2003). The second device was somewhat simpler and smaller with a
single-valve system, a single flow path for inhalation and exhalation and a single
flow monitor. Patients undergo a training session to determine the individual
tolerance to breath-hold and the length of time for which they can comfortably
349
Figure 6.50. A schematic representation of the active breathing control (ABC) apparatus
showing a single-valve system with a single flow path interfaced to a personal computer.
(Reprinted from Stromberg et al 2000 with copyright permission from Elsevier.)
hold their breath. They also view a real-time display of the changing lung volume
as well as the intended breath-hold level. Once this is individually established, the
ABC device is used to halt the patients breathing for reproducibly fixed periods
in the breathing cycle.
The technique of ABC has been used to study the reduction in normal
tissue complications with deep inspiration when treating patients for Hodgkins
disease (Stromberg et al 2000) (figure 6.50). The clinical goal was to attempt
to reduce the dose to the normal lung and to the heart without compromising
the treatment of the tumour. This is desirable given that patients generally
survive Hodgins disease but can then progress to suffer from late normal-tissue
radiation complications including secondary cancers, cardiovasular disease and
pulmonary dysfunction. ABC may also reduce cardiac overlap by moving the
spleen inferiorly with deep inspiration.
Five patients underwent CT scanning of the chest and abdomen using the
ABC apparatus. Scans were taken at normal inspiration (NI), normal expiration
(NE) and deep inspiration (DI) in the supine position. DI scans were also repeated
during and at subsequent weeks to assess the intrafraction and interfraction
reproducibility. Stromberg et al (2000) described the CTVs and PTVs defined
at NI, NE and DI. A composite PTV was also defined based on the range of NI
and NE CTVs. Planning on an ADAC PINNACLE system used conventional AP
and PA beams.
The assessment of normal-tissue damage was made using dose-volume
histograms for the heart and dose-mass histograms for the lung, the latter to
remove the effect of changing air volume at different stages of the breathing
cycle (which is irrelevant in calculating tissue damage). These dose-volume/mass
histograms were compared at the several respiratory phases.
ABC was found to be well tolerated with DI breath-holds ranging from 34
to 45s. Dose-mass histograms for all five patients showed a median reduction of
12% lung mass irradiated at DI compared with free-breathing. Advantages were
also found for breath-hold using ABC at NI and NE compared with free breathing
350
but these were less significant. All patients showed a decrease in heartspleen
overlap with DI of at least 1 cm. Cardiac DVHs showed the mean volume of heart
irradiated at the 30 Gy dose level decreased from 26% to 5% with DI compared
with free breathing. Analysis showed intra- and inter-session variabilities of only
5%. It was concluded that all ABC-controlled treatments led to improvements
with DI providing the greatest benefits.
Aznar et al (2000) have used the ABC device in conjunction with IMRT for
improving the sparing of heart in left-breast radiotherapy. Remouchamps et al
(2002, 2003) have also shown how the use of deep-inspiration breath-hold ABC
can lead to lower cardiac doses for left-breast-irradiated patients.
Sixel et al (2001) have investigated the use of deep-inspiration breath-hold
during tangential breast radiation therapy as a means to reduce cardiac irradiated
volume. The ABC device was used for five left-sided breast cancer patients.
It was found that the use of the ABC device reduced the dose to the heart for
some patients when deep-inspiration breath-hold was applied. The magnitude
of the impact of the breath-hold technique depended on the patient anatomy, lung
capacity and pulmonary function. This was determined by creating plans based on
CT scans which were acquired with and without the breath-hold and subsequent
virtual simulation for regular tangent and for wide tangent techniques.
Donovan et al (2002a, 2003) has used the ABC device for four patients
receiving tangential pair radiotherapy to the left breast. Patients are able to
achieve a 1520 s breath-hold for 68 repeat breath-holds with 75% of the
maximum lung capacity and the heart was excluded from the treatment fields
in all cases. The OSIRIS system was used to acquire outline information with
and without the ABC to assess shape changes and reproducibility of position. At
the same centre, McNair et al (2003a) have measured the mean breathold time
as 20 s (range 1025 s) for ten patients with lung cancer. The mean breath-hold
levels were 72% of the maximum inspiration volume. Christian et al (2003a)
reported good tolerance of the device. A review of the application of ABC at this
centre is in Wavelength (2003d).
Wilson et al (2001) have used the ABC device for treating non-small cell
lung cancer with CHARTWEL. Five patients with locally advanced NSCLC
underwent three consecutive planning CT scans, one breathing, and the second
and third using ABC. ABC was found to be tolerated for breath-holds of between
2030 s and the CT lung volumes were reproducible. PTVs could be made smaller
using ABC with the median reduction of approximately 50 cm3 . Wilson et al
(2004) reported on a longer study and showed that the use of the ABC device also
led to increased lung and spinal cord sparing.
Dawson et al (2000) have assessed the use of the ABC device for improving
the treatment of liver cancer. Over 100 fractions of radiation have been delivered
using ABC in four patients. The average time for each breath hold was 25 s in a
range 1045 s. The motion of the diaphragm and hepatic markers were detected
on fluoroscopy during the ABC breathhold and found to be reproducible to within
3 mm over the short term. The average long-term reproducibility of the diaphragm
351
position increased to about 6 mm indicating a drift of breath-hold liver position

during the course of treatment. The technique relied on imaging liver micro coils,
which have been inserted during hepatic artery catheter placement.
A company perspective on the ABC (C for control has become C for
coordinator) device has been written (Wavelength 2002a). A description of
the clinical implementation in two centres (William Beaumont Hospital and
Mount Vernon Hospital) appeared in Wavelength (2002b) and at Thomas Jefferson
University (TJU) and the Royal Marsden NHS Trust in Wavelength (2003d).
Wong et al (2001) have shown how a combination of ABC and respiratory gating
can offer an improved approach to target volume immobilization during treatment
using an Elekta machine. The TJU DAO method is incorporated in the Elekta
PrecisePLAN treatment-planning system (Wavelength 2002c). The use of the
ABC device for guiding extracranial stereotactic radioablation is described in
Wavelength (2002d).
6.10.12 Calculating the effect of tissue movement
6.10.12.1 Incorporating movement knowledge into the inverse planning itself
Li and Xing (2000a, b) have discussed the problem of accounting for random
organ movement in IMRT. Firstly, they proposed a method to assess the outcome
of movement applied to a plan computed without accounting for tissue movement.
Call this plan Df (n) where n labels patient voxel and f represents fixed anatomy
during inverse planning. If voxel n has probability P(n, n ) of moving to voxel
n , then the dose received by voxel n is

P(n, n )Df (n ).
(6.7)
D(n) =
n
The dose-volume histogram computed from D(n) then gives the outcome of
incorporating organ movement into the plan computed without organ movement.
A better approach, however, is also suggested in which the objective function
includes the effects of movement, i.e.
obj =
N
1
rs [D(n) D0 (n)]2
N
(6.8)
n=1
where D0 (n) is the prescribed dose and rs the IF for structure s and N is the
total number of dose-calculation points. Minimizing this cost function minimizes
the difference between the motion-smoothed dose distribution (from equation
(6.7)) and the prescription. The algorithm was incorporated within the PLUNC
treatment-planning system and the simultaneous iterative treatment planning
(SITP) iterative algorithm was used for cost-function minimization. Li and Xing
(2000a, b) applied this to two clinical cases describing the motion using a 3D
Gaussian distribution function. It was shown that the second approach gave better
critical-organ sparing than the first for much the same PTV coverage (figure 6.51).
352
Figure 6.51. The CTV, rectum and bladder DVHs in two different cases: case 1, the
DVHs of the treatment plan taking into account the spatial uncertainties of the structures
during the optimization with the method described (full lines); case 2, the DVHs of the
treatment plan by requiring a uniform dose to the PTV (CTV plus 1 cm margin) and zero
dose elsewhere (broken lines). (From Li and Xing 2000a.)
Xing et al (2000b) have investigated the effect of respiratory motion in the

design of IMRT treatment for breast cancer. The PLUNC treatment-planning
system was used. It was found that if the effects of motion of the breast were not
included in the static plan and then this plan were applied to geometry in which
the breast motion was switched on, the dose received by the breast differed
significantly from what had been planned in the static situation. For example,
the minimum target dose to the CTV decreased by between 69% for all six
cases studied and the volume receiving lower doses increased. As a result, the
average dose in the clinical target volume was decreased significantly. However,
if the motion were included in the inverse treatment planning, then the plans
obtained with that approach improved the situation greatly and represented the
truly optimal solution in the presence of the respiratory movement of the breast.
Loof et al (2001) have stressed the importance of including geometrical
uncertainties in the optimization process for IMRT. These were simulated by
convolving a treatment plan with the probability function representing expected
movement. An alternative approach was to create IMRT plans with the HELIOS
treatment-planning system in which the pencil-beam kernels had already been
convolved with a movement kernel. The two approaches were expected to be
similar. The authors then went on to point out that by using the second of these
techniques ab initio the inverse planning could automatically take account of the
motion of tissues.
6.10.12.2 Use of multiple CT datasets and adaptive IMRT
Using multiple CT scans to assess a plan made from a day-1 scan
Xu et al (2000) have performed multiple daily CT scans for 30 prostate patients
with an average of 17 scans per patient. These scans have been registered
using the bony anatomy. The plan produced for IMRT of the prostate on day
1 was then reapplied to the geometry obtained from the subsequent CT scanning
353
occasions and, in particular, the change of dose to the rectum was evaluated. It
was found that large dose-rectal-wall volume variations existed between patients
whereas, by contrast, the dose-rectal-wall volume variation for any one patient
was somewhat smaller. It was found that making use of the first few consecutive
CT images improved the quality of treatment planning for rectal-toxicity-based
dose escalation.
Bignardi et al (2000) have studied the effect of variability in the PTV over a
period of fractionated treatment of the prostate. Eighteen cases were studied, each
with a four-field box technique for CFRT. The CT scans were performed initially
and at three and six weeks of treatment and the PTV was re-computed at three
and six weeks from these subsequent scans. The beam parameters from the plan
using the initial CT scan were then reapplied to the data obtained at three and six
weeks. The complication probabilities for the target volume and for the rectum
and bladder were remarkedly constant over the six-week period.
Large et al (2001) have studied the effects of interfraction prostate motion on
IMRT delivery. A series of CT scans was taken in the supine position immediately
before and after the first three fractions. These scans were then registered. The
targets were redrawn and the targets shifts were assessed. The dosimetric effects
were seen to be as large as 10%.
Plasswilm et al (2002) made multiple CT scans of patients receiving IMRT
for prostate cancer and showed that in two out of nine patients there would have
been significant underdosage of the CTV if the plan from just the first CT scan
rather than the plan from some averaged CT scan had been used.
Happersett et al (2003) have performed a study of the effects of internal
organ motion on dose escalation in conformal prostate treatments. This study was
performed for 20 patients. For each of the patients, a planning CT dataset was
created and plans were made on this dataset. Three types of plan were made;
the first was a conventional six-field plan with a prescribed dose of 75.6 Gy; the
second was the same six-field plan escalated to 72 Gy followed by a boost to
81 Gy and the third was a five-field plan with IMRT delivering 81 Gy. For each
of these patients and for each of these three types of plan, three more subsequent
CT scans called treatment CT scans were made. The scans were made on the
first day of treatment, at the fourth week and at the completion of the treatment
course. Each of the three treatment CT scans was then spatially registered in 3D
to the planning scan using a chamfer-matching algorithm and the contours from
the treatment scans were transferred to the planning scan. It was then possible
to evaluate the planned dose distribution on contours representing organs in the
treatment scan. No attempt was made to separate organ motion into systematic
and random components and in the dose calculation, for what is known as the
treatment results, rather than planned results, the dose from each treatment scan
was assumed to contribute one-third of the total dose for the entire treatment (this
may not be too representative). The study showed that for the planning scans, the
TCP for plan types 2 and 3 was 9.8% greater than for plantype 1. Conversely
for the treatment TCP, the increases were 9% between plans 2 and 1 and 8.1%
354
between plans 3 and 1. It was thus observed that the IMRT plan was more
susceptible to patient movement. However, given that the TCP for the planned
IMRT treatment was already higher than that for the other two treatments, dose
escalation with IMRT was deemed to be possible. This is a good example of the
use of multiple CT scans to assess the robustness of specific treatment plans to
organ motion.
Hoogeman et al (2003) have used multiple CT scans (a planning CT scan
and 11 repeat CT scans) to determine interfraction prostate motion and, from this,
have been able to reduce the systematic error in IMRT of the prostate.
Schaly et al (2004) have developed a technique to actually track the dose to
the whole 3D volume when the movement of individual voxels is computed from
a series of CT scans. The William Beaumont CT datasets were used (15 CT scans
for the same patient) and a thin-plate spline method was developed to create a
map of the movement of all the tissue voxels. The application was to conformal
(not IM) radiotherapy of the prostate. It was found that large (typically 30%)
changes to the dose to the rectum occurred for the first fraction and that even after
summing over all 15 fractions the rectal dose was still some 20% different from
that planned. The prostate dose did not vary much. This is a very significant study
since it is one of the first to actually track the voxel motion and compute effects.
Using multiple CT scans to create a composite target volume
McShan et al (2001, 2002) have developed the multiple instance geometry
approximation (MIGA) which includes defining two or more instances of the
patient geometry and optimizing the plan for all instances concurrently. This
is a way of taking account of, at present, rigid body translations with respect to
one particular planning geometry. Errors due to intrafraction movement are not
included. The MIGA solutions are worse than the no motion inverse planning
but better than the latter convolved with motion. Fraass et al (2002) have used
MIGA to improve IMRT for head-and-neck CTVs which are unusually near the
skin surface.
Martinez et al (2001) have described an adaptive radiotherapy technique
(ART) to customise margins of expansion from CTV to PTV for the individual
patient. CT and portal images were recorded throughout week 1 of treatment
and used to define a confidence-limited PTV in which anisotropic margins were
applied. The confidence-limited PTV, PTVCL , was some 24% smaller on average
than the conventional PTV. The achievable level of dose escalation is patient
specific and, on average, it was found to be 7.5% for IMRT.
Pavel et al (2001), and Pavel-Mititean et al (2004) have studied the
movement of the prostate and rectum during a course of treatment using nine sets
of CT scans taken at intervals throughout the treatment for a series of prostate
cases. The volumes of target, bladder and rectum were calculated and compared
with the planning-scan volumes and, in particular, dose-volume histograms and
dose-surface histograms for the mobile rectum were calculated and compared
355
for plans made on each CT scan. A total dose-surface histogram was then
compared with the daily dose-surface histogram. It was found that the target
volume changed much less than the changed volume of bladder and rectum. The
largest translations were seen in the craniocaudal direction. When comparing
dose-surface histograms from individual fractions with the planning case, the
percentage of the OAR (rectum) surface receiving 6080% of the target dose
increased from 6068% for conventional plans and from between 3045% to 40
55% for IMRT plans.
Regarding time trends in prostate movement, Mechalakos et al (2002) have
reported on a series of 50 patients each of whom had four CT scans during their
course of treatment. The study investigated the time course of a large number
of physical parameters associated with the planning. It was found that only two
parameters showed significant changes, namely the total bladder volume (despite
controls) and the increasing separation between bowel and PTV. No trends were
observed for the prostate, seminal vesicles or rectum. This study had all patients
prone. Its conclusions imply that repeat scanning of patients is not needed,
somewhat at odds with other studies.
Using multiple CT scans to change the inverse-planning technique
Errors in the set-up of patients for radiotherapy, like all errors, can be both
systematic and random. To estimate the systematic error, one has to take several
measurements during the first few fractions of (say a 30 fraction) treatment.
Bortfeld et al (2002d) have mathematically analysed and solved the problem of
determining at which fraction an intervention should take place to minimize the
overall systematic error. The idea is that for the first n treatment fractions the
error is measured. At the (n + 1)th fraction, a correction is made, being the
difference between the actual position of the target and the mean of the errors
so far and this correction is then applied for all further fractions up to the total
number of fractions. The question is what is the value of n? The authors wrote
down expressions for the expectation value of the overall quadratic set-up error
and then minimized that function. Making the assumption that both the actual and
the individual positions are parts of distributions which can be arbitrary, with two
different variances, they found that for 30 fractions and typical values of these
variances the curve reached a minimum at about n = 4. The curve was very
flat-bottomed indicating that the exact fraction of intervention was probably not
critical.
Wu et al (2002a) have proposed a novel technique in which errors introduced
into the delivery of the up to nth fraction can be corrected by re-optimizing
the (n + 1)th fraction (and so on for all n). The method relies on repeated
measurement of patient position and dose reconstruction at each fraction.
Essentially the (n + 1)th optimization is of the total dose prescription minus the
cumulative dose to the (n + 1)th fraction. Several alternative strategies were
proposed. However, the technique assumes no deformable patient. The most
356
appropriate apparatus capable of implementing this is the University of Wisconsin

Tomotherapy Unit. Wu et al (2004) propose that the Cimmino algorithm can be
used to rapidly compute the dose distribution for a warped geometry using the
starting parameters from the unwarped geometry.
Birkner et al (2001, 2002) have incorporated image feedback into inverse
planning. Patients with prostate cancer were sequentially CT scanned during their
treatment and a model was made of daily set-up errors and organ motion and
deformation. The information from CT scans over the first five days was fed back
into the inverse treatment-planning process. Birkner et al (2001) incorporated
organ motion and deformation into inverse planning (4D IGRT). The rectum is
known to move during a course of treatment and motion-averaged dose-volume
histograms were developed using a series of CT scans taken through the treatment
time. These were then used for optimization. It was claimed that this reduces the
high dose to the rectum. The future intention is to provide a seamless integration
of HYPERION to ADAC PINNACLE.
Birkner et al (2003) have described on offline method to perform adaptive
radiotherapy to account for variations in set-up and internal organ motion
throughout the treatment fractions. Planning was performed on an initial CT
dataset regarded as a reference situation for the geometry. Then (m 1) further
CT datasets are acquired. These m datasets were used to estimate the positions
of all voxels in all structures as a fraction of time represented by a probability
fraction for the tissue location and this plan was then applied for the remaining
(n m) fractions. The final outcome depends on the value of m but Birkner et al
(2003) showed that m = 5 gave a more conformal overall result (than m = 1) and
that continuous adaptation beyond m = 5 made marginal difference.
Didinger et al (2001) have shown the advantages of using adaptive inverse
planning during IMRT of prostate cancer. Weekly CT scans were used to replan
the patients.
Daily repositioning based on daily CT scanning in the treatment room
Kneschaurek et al (2000) have generated daily CT images prior to irradiation
of patients with prostatic cancer. These images were then transferred to the
BrainSCAN treatment-planning system which automatically generated optimal
leaf positions for the BrainSCAN microMLC. By registering the daily CT data
sets with the first CT investigation, it was possible to generate a composite dosevolume histogram for the whole treatment, which took account of the movement
and different filling of the rectum.
Ruchala et al (2002a) have presented a different way to reposition the
patient on any given fraction that does not use external fiducial or bony anatomy
repositioning. This is called contoured anatomy dose repositioning (CADR).
Each time a new CT scan is created, the relevant targets and OARs are contoured.
These contours are then compared with those for the planning CT images and
patient repositioning is based on the use of these contours. Ruchala et al (2002a)
357
Figure 6.52. A panoramic view of the integrated CTLinac irradiation system. A linac
gantry is placed on the opposite side of the CT gantry. Between these gantries, the common
treatment couch is placed. (Reprinted from Kuriyama et al 2003 with copyright permission
from Elsevier.)
describe the advent of onboard 3D imaging for radiotherapy, which promises

to vastly increase knowledge of patient anatomy at the time of treatment as
an example of a conundrum called Cassandras complex of Greek mythology.
Cassandra was granted knowledge of the future yet was powerless to change it.
The CADR technique is a first step towards optimizing this problem.
Kuriyama et al (2003) have designed an integrated facility with the CT
scanner in the same room as the linac. The CT scanner moves on rails relative
to the patient on the couch rather than the more usual vice versa (figure 6.52).
The centre third rail carries a magnetic strip with position information ensuring
the patient position is correct to better than 0.4 mm for stereotatic radiotherapy
(figure 6.53). The arrangement also permits verification of the position of targets
for CFRT. Paskalev et al (2003, 2004) have designed a technique to correlate
such in-treatment-room CT images with planning CT images and to generate the
required shifts to align the two on a daily basis. The correlation is based on target
information only; no contours need to be defined. It was shown that the method
is robust to noise and as good as a method based on image fusion. Reported shifts
in the position of the prostate were up to 9 mm.
358
Figure 6.53. Diagram of the integrated CT-Linac irradiation system. The treatment couch
has two rotation axes: A1 is for isocentric rotation to make non-coplanar arcs; and A2 is for
the rotation between CT and linac. The gantry axis of the linac is coaxial with that of the
CT scanner. A special self-driving device is equipped on the bottom of the conventional CT
gantry so that the gantry, not the couch, moves when scanning. (Reprinted from Kuriyama
et al 2003 with copyright permission from Elsevier.)
Dong et al (2004) have used the Varian ExaCT CT-on-rails system coupled
to a Varian 2100 EX accelerator for in-treatment-room repositioning of the patient
at each fraction. By deliberately creating shifts, it was found that the system could
correct them to better than 1 mm. The system was used to show, and correct for,
GTV shrinkage through a course of treatment as well as to correct interfraction
variations in position.
Progress in MVCT and kVCT will be reviewed in section 6.11.
Adapting the delivery of IMBs to take account of interfraction motion
Ruchala et al (2002c) have introduced multi-margin optimization with daily
selection for image-guided radiotherapy. The study was conducted retrospectively
using 17 daily CT scans acquired during a prostate treatment . PTVs were
identified with a variety of margins and IMRT were generated for these PTVs
(four of them). For each fraction, the most appropriate plan was chosen
corresponding to the daily position of the prostate. It was shown that this led
to decreased rectal-and-bladder dose.
359
6.10.12.3 Modelling the effect of intrafraction movement

CT scans at different acquisition speeds and at different phases of the breathing
cycle
Because lung tumours move during radiotherapy with respect to the beams
applied, the determination of lung mobility is a key research area. Van Sornsen
de Koste et al (2003b) have made a study of this subject by collecting, for each
patient with non-small-cell lung cancer, three rapid and three slow planning CT
scans. They then computed a so-called optimum CTV as the envelope of the
CTVs determined for all of these six scans. The six scans were co-registered using
contour-matching software ACQSIM. The relationship was studied between the
CTVs determined from the slow scans, the rapid scans and the optimum scan and,
not surprisingly, it was found that this ratio was less than one. The ratio had a
mean of 0.68 for the rapid-to-optimal CTV and 0.78 for the slow-to-optimal CTV,
thus indicating that the position of the CTV as determined from the mean of the
slow CT scans was closer to that regarded as optimal. The key observation was
that by applying a standard margin of 5 mm symmetrically to the slow-scan CTV,
the optimal CTV would be covered. There was no correlation, however, between
tumour mobility and tumour location. They thus concluded that individualized
procedures were required for each patient.
Allen et al (2004) performed a similar study. Three CT scans were made
at end expiration, end inspiration and free breathing. GTVs were defined on
each scan. It was found that the application of a margin to the GTV on the freebreathing scan did not adequately cover the tumour position as determined from
the end tidal scans. In some cases, areas of tumour were geographically missed
and, in other cases, too much normal lung was included.
Keall et al (2004) have coupled the Varian RTM infrared system to a 16-slice
helical CT scanner to organize projections according to phase in the breathing
cycle. Since the CT scanner was expecting a cardiac pulse for triggering, the
pulses from the infrared system needed to be shortened for this gating procedure.
Then images were reconstructed of different phases. This demonstrated the
system worked well for a moving phantom and also for a patient undergoing
audio-visual breathing coaching during scan acquisition. Images of the moving
lung tissue were presented. However, when the breathing pattern was irregular,
missing projection data led to reconstruction artefacts. It was also demonstrated
that slow CT scans could be reconstructed by ignoring the phase tag and that
the slow scan was equivalent to the scan at any individual phase of a stationary
phantom.
Kim et al (2003) performed experiments to show that the definition of target
was too small when the total time of the CT scan was less than one breath time.
The target was correctly defined with a CT scan of exactly one breath period and
did not change if the CT scan time thereafter increased.
Frazier et al (2000, 2004) have evaluated the impact of respiration on wholebreast radiotherapy using either wedged pairs or IMRT at the William Beaumont
360
Hospital. CT scans were taken with the patient breathing normally, a so-called
free breathing (FB) scan and, at the same time, the ABC apparatus was used
to obtain two additional CT scans with the breath held at the end of normal
inspiration (NI) and normal exhalation (NE). The plan produced using the FB scan
was then copied and applied to the NI and NE scans. Somewhat surprisingly, the
dose distribution to the whole breast was observed to change very little between
the FB, NI and NE scans for both the wedge technique and IMRT indicating that
the dose distributions delivered using multiple-static MLC segments are relatively
insensitive to the effects of breast motion during normal respiration. However,
this is not true when considering the nodes which may be designed to be excluded
during the FB irradiation but turn out to receive significant dose when the same
plan is applied to the CT scans taken at normal inspiration and exhalation.
Studies of the effect of movement for a single fraction
Holmberg et al (2000) have studied the effect of target movement in step-andshoot IMRT via numerical simulation. The intrafraction position of a target
element in the thorax region was modelled as a function of time through an
asymmetric cosine function with a bias towards the exhale position. It was
found that the maximum simulated exposure time in a moving target was as
high as 170% of the predicted exposure time in a static target when the shorttime segment covers the inhalation position and the long-time segment covers the
exhalation position. With opposite positions, the calculated maximum reduced
to 130%. However, this demonstrates that intrafraction movement can seriously
affect IMRT of the thorax for a single fraction of IMRT.
Kung et al (2000c) and Zygmanski et al (2001d) have developed a technique
to predict the dose error in IMRT treatment of lung cancer due to lack of account
of respiratory movement. The IMBs from an inverse-planning system were
converted into leaf positions for the dMLC device at each portal. Then, for the
beams-eye view of each portal, a 2D fluence grid was created and applied to
the moving target volume. This enabled the time-dependent sub-fields to be rayprojected to create an effective incident fluence. This effective incident fluence
was then returned to the Varian HELIOS treatment-planning system to produce
the dose distribution that is actually received by the moving target at a single
fraction. It was found that, without respiratory gating, target volumes could
receive dose errors of up to 10% or greater over 15% of the target volume pointing
to the need for respiratory gating in thoracic IMRT.
Ramsey et al (2001b) have considered that the dosimetric consequences of
intrafraction prostate motion in IMRT could be disastrous. They argued that
IMRT only works by lining up very precisely different modulated fields and,
if the prostate moves between these fields, then the dosimetry will go wrong.
Ramsey (2002) has experimentally simulated the effects of respiration motion
by driving the patient couch in a simulated respiratory motion relative to the
delivery of IMRT using the dMLC technique. Using film dosimetry, they found
361
differences between the dose distribution under these circumstances compared

with that when the couch was static (i.e. simulating no motion). With respect to
the studies of Bortfeld et al (2002a, b), it must be noted that this is simulating
only a single fraction delivery and the effects should be less when averaged over
a set of fractions (see next section).
Sohn et al (2001b) have studied the effect of motion on lung-cancer IMRT.
Often the effect of motion on a dose distribution is modelled by convolving
the static-patient dose distribution with the mathematical representation of
the movement. Sohn et al (2001b) did this for a specific modulated field
created by CORVUS. They then instead created an experiment in which they
measured the dose distribution created by a modulated field when the detector
was on an oscillating frame simulating motion. Interestingly, they found that
this measurement differed from the convolution calculation by up to 15%.
The discrepancy was nearly as large as the motion-induced artifact itself. It
was thus concluded that motion convolution models inadequately predict the
breathing-motion-induced dose fluctuations from IMRT and a more sophisticated
model of the artefacts caused by breathing as a function of the radiation dose
delivery sequence is warranted to enable development of motion-robust delivery
algorithms. This observation does not disagree with the conclusions from Bortfeld
et al (2002a, b) because it is a one-fraction experiment and Bortfeld et al
(2002a, b) explained that individual fraction dose distributions can differ from
calculation. The dose distribution averaged over many fractions should not (see
next section).
Pemler et al (2001) have studied the influence of respiration-induced organ
motion on dose distributions in treatments using enhanced dynamic wedges. They
studied the particular situation in which craniocaudal movement of a tumour
began to beat with similar craniocaudal movement of a dynamic wedge for a
Varian accelerator. In worse cases studied, the maximum deviation in MUs was
16% for a 60 wedge. It was commented that similar disagreements might result
for IMRT with the dMLC technique.
Schaefer et al (2004) delivered five modulated fields, corresponding to a
lung irradiation, to a moving phantom with the oscillation period of either 12
cycles min1 or 16 cycles min1 and an effective amplitude of 16 mm. The dose
was measured at 18 points and compared with that received by a static phantom.
The delivery technique was the static step-and-shoot (MSF) method. It was found
that, even for a single fraction, the dose varied by no more than 5% when motion
was occurring. It was therefore concluded that the damage due to motion would
be even less with fractionated radiation therapy (see next section).
Studies of the effect of movement for the whole radiotherapy course
Bortfeld et al (2002a, b) have investigated the effects of intrafraction movement
for IMRT delivered with a compensator, with an MLC and with a scanning beam
(figure 6.54). A sinusoidal motion was modelled with an amplitude of 5 mm
362
and fractionation was considered. To calculate the effect of movement, for the
compensator the dose matrix was simply moved with respect to the compensator.
For the MLC IMRT technique, the individual field patterns were delivered to
the moving dose grid and similarly for the scanning beam. It was found that
the expected delivered dose distribution was the convolution of the movement
pattern with the distribution without motion, i.e. exactly the same as observed for
conventional radiotherapy. So the effect of movement could be accommodated by
the use of margins, as with conventional therapy. Additional effects specific to the
IMRT delivery technique were small. This conclusion is largely as a result of the
averaging of individual daily movement patterns over the complete course of (say)
30 fractions. Errors due to organ movement on a given day may be quite large but
these will be averaged out over the course of the fractions and were not expected
to have biological consequences. It was shown from theoretical considerations
that the probability function for the relative deviation of dose from the expected
value quickly becomes Gaussian as the number of fractions increases. The width
of the Gaussian depends on the IMRT technique. However, this does not mean
that attempts to reduce the effects of motion are wasted since margins are never
a wanted phenomenon. They compromise conformality to CTV and any attempt
to reduce them by the use of gating, tracking etc would be useful. The study
was made assuming the dose distributions do not deform with breathing but just
relocate relative to the modulation (see also extensions discussed by Bortfeld et
al [2004]). Chang et al (2004b) disagree with the conclusions of Bortfeld et al
(2002) and argue that deformable organ movement will cause a greater interplay
between time-dependent delivery of IMRT to moving tissues and their movement.
Jiang et al (2002, 2003c) conducted experiments irradiating a motor-driven
platform containing a phantom by different IMRT techniques. The platform
oscillated with an amplitude of 2 cm and a period of 4 s to mimic the motion
of a lung tumour to first order as a sinusoid. Measurements were made at eight
positions in the breathing cycle, at two dose rates, for five fields from two patient
plans and for a dMLC delivery and MSF delivery with either 10 or 20 intensity
levels. The phase of the movement with respect to the start of delivery was varied.
Hence, the experiment observed an interplay effect.
Not surprisingly large dose variations (as large as 30%) were found for
delivery of single fractions, but the FWHM of dose dramatically narrowed (to
about 2%) when the average was taken over 30 treatment fractions. In the study of
Jiang et al (2003c), the direction of movement was always at right angles to that of
the leaf movement whereas in the study of Jiang et al (2002) four angles between
MLC and phantom motions were simulated. No differences were observed
between delivery techniques. There was no apparent dependence of interplay
between the dMLC delivery mode or the angle between MLC motion and tumour
motion. This is good news for conventional lung IMRT but somewhat perversely
may be bad news for gated IMRT in which the delivery and the movement of
the tumour is highly correlated. Also, despite the conclusions, it was recognized
363
Figure 6.54. Illustration of the interplay between organ motion and leaf motion for the
delivery of IMRT with an MLC. The leaves move from left to right. The star symbolizes a
point in an organ that moves up and down. The two different versions of the star represent
two different phases of the motion. (Let us say that the filled star represents exhalation and
the open star represents inhalation.) Depending on the phase relative to the leaf motion, the
point can receive very different dose values. In the phase shown by the filled star, the point
dose not receive any primary dose between the time t1 and t4 and it may, in fact, receive
no primary dose at all. In the phase symbolized by the open star, in which the point moves
up between t1 and t4 , it is treated with the full primary dose at all times. This example,
of course, represents the extreme limit of possible dose variation and such extremes are
certainly relatively improbable. (From Bortfeld et al 2002a.)
that modelling is limited and that tumour tracking will be preferable for CTV
conformality.
Duan et al (2002) have delivered IMRT to a phantom which was set
into linear sinusoidal motion with 1.5 cm longitudinal and 0.5 cm transverse
amplitude to simulate respiratory motion. Dose-volume histograms show marked
differences after one fraction, maxima and minima varying up to 20% from what
was required. However, when five fractions were delivered with de-phasing
between the fractions, the absolute errors decreased to about 5%. These results
demonstrate that although respiratory motion may introduce substantial dose
errors in a single treatment, multi-fraction treatments can significantly reduce
these errors.
Kim et al (2003, 2004a) have computed the time-averaged dose-volume
histogram (TDVH) for a treatment by convolving the dose distribution with
the probability function for movement of targets and OARs. For a head-andneck case, they showed that, with a standard deviation of movement of only a
millimetre or so, the time-averaged DVH is not significantly different from the
non-time-averaged DVH for the PTV although there are differences for the OAR.
Chui et al (2003a, b) explained how intrafraction organ motion interferes
with the delivery of IMRT by the dMLC or shuttling MLC technique. If the
motion is along the direction of leaf travel, then the primary fluence will not
be calculated by a vertical-line intersection in the timedistance diagram of
364
leading and trailing leaf patterns but instead by the intersection of the motion
line with this diagram. If the motion is perpendicular to the leaf motion, the
irradiation points may sample modulations from the adjacent leaves (figure 6.54).
Chui et al (2003b) modelled the movement of targets using a function whose
amplitude varied from 3.5 to 10 mm and whose period varied from 4 to 8 s.
IMRT planning for three breast patients and four lung patients was considered.
Calculations were made for each case with 30 randomly chosen phases and
fluence maps were then calculated as the sum of these overall 30 phases to
simulate the effect over the whole fractionated radiotherapy. Dose distributions
were computed with these whole course averaged fluence patterns. This whole
procedure was repeated 10 times to estimate uncertainties. For breast patients
(movement parallel to leaf movement), the statistics showed small variations
about the mean attributed to the fact that the delivery averages over a very large
number of periods. The high-dose portion of the DVH does not vary by more than
2% from the result with no motion. The low-dose portion is more affected. This
implies the margins were sufficiently large that the PTV coverage was not unduly
affected by motion. Conventional treatments (non-IMRT) behaved the same. Very
similar results were obtained for the lung patients (movement-perpendicular to
leaf movement) except that the low-dose parts of the PTV were more greatly
affected. However, the CTV statistics showed almost identical coverage to the
calculations without movement. If the margin cannot be increased, then some
other intervention such as breath-hold or ABC must be used.
Rosu et al (2003) have estimated the effect on liver target and normaltissue dose due to set-up error and organ motion relative to a planned dose
distribution. Forty patients were studied who had had pre-treatment CT scans
taken with voluntary breath-hold at normal exhalation. The effects of two forms
of motion were computed using convolution techniques. The first form of motion
was interfraction set-up variation which they represented using an anisotropic
Gaussian kernel to describe the set-up uncertainties. The static dose distribution
was convolved with the appropriate kernel to result in a new dose distribution that
directly incorporates the effect of geometric uncertainties. Then this blurred dose
distribution was further convolved with a 1D (superior-inferior direction only)
patient-specific probability distribution function to describe breathing in which
more time is spent at exhalation than at inhalation as described by the probability
function
1
(6.9)
p(z) =

z0 z 2n1
z0 z 1/n 0.5
2n
nb b
1 b
where terms are defined after equation (6.6). For the double-convolved dose plans,
dose-volume histograms and recomputed normal tissue complication probabilities
at the prescription dose were compiled. Also the prescription dose in the static
plan was re-calculated at a nominal 20% NTCP level and then the dose required
in the convolved plan to restore the NTCP level to 20% was re-evaluated.
365
As expected, the convolution led to blurred dose distributions creating

penumbra regions at beam edges. It was found that the corrections due to
breathing predominated over the set-up variations and the main result was that, as
a result of breathing, hot spots emerged superior to the structures of interest and
cold spots emerged inferior to the target volume compared with the static dose
distribution. The minimum doses to the CTVs met or exceeded the minimum
PTV doses from the static plans in all but one case. Detailed dose-difference
distributions and tables of NTCP differences in Rosu et al (2003) need to be
studied to understand the detailed changes in dose and complication probability.
Finally, it was commented that convolution assumes the shape of the dose
distribution is unaffected under small translations of the patient geometry and
also assumes no change in the external contours or deformation of the organs. All
of these assumptions are strictly untrue and work is continuing to develop a model
of tissue deformation.
Modifying the IMBs to account for motion
Deng et al (2001c) have incorporated breast organ motion into the MLC leaf
sequencing for IMRT. The leaf sequencer reads in the breathing pattern as a
function of leaf-pair positions as if the breathing pattern were fed into the MLC
controller. The MLC leaf trajectories were then adjusted based on the breathing
motion and MC dose calculations were performed using MCDOSE at different
phases of the breathing cycle. The inclusion of motion into the adaptation of
the IMRT delivery effectively delivered a plan equivalent to that without organmotion correction.
Deng et al (2002) have investigated the importance of breathing motion on
IMRT treatment. They incorporated the effects of organ motion into the MLC
leaf sequencing. When these components were applied to the moving patient, the
dose distribution was more accurate than when the corresponding sequence, not
accounting for movement, was applied to the moving patient.
Neicu et al (2002, 2003) have developed gated motion adaptive therapy. The
original IMRT leaf sequence is modified to compensate for tumour motion based
on the average tumour trajectory. The average tumour trajectory was derived by
averaging tumour position at each phase of breathing. The radiation beam moves
along the average tumour trajectory while the tumour moves according to the
measured data.
Gierga and Jiang (2002) and Gierga et al (2003) have evaluated the effect
of target motion in IMRT. Measured tumour trajectory data for lung patients
were used and IMRT plans were created with a commercial IMRT treatmentplanning system. A sampling algorithm was then used to generate the photon
fluence maps that incorporate tumour motion using the original leaf sequence
files and knowledge of the tumour trajectory. These modified fluence maps
were then input to a MC dose calculation to calculate the 3D dose distribution.
Different initial phases in the breathing cycle were simulated for 20 fluence maps
366
and the corresponding dose distributions for 20 equally spaced initial phases
were computed. The expectation value and variation of point doses and dosevolume histograms were then estimated and found to be different for each of the
phases. Individual dose distributions could differ widely from those planned. The
overall dose distribution was calculated by sampling randomly from the initial
dose distributions for each phase of all fields. A distribution averaged over the
phases differed much less from that planned. Gierga et al (2004) performed a
similar study for liver tumours. Fluoroscopy of clips established the trajectories
and IMBs for static plans were then re-applied to the moving targets. Perhaps
surprisingly, it was found that liver motion did not greatly affect the distributions
for many of the patients studied. Jiang et al (2003a, b) showed how to make a
SMART (Synchronized moving aperture radiation therapy) plan for IMRT. Fourdimensional CT data (Low et al 2004, Wahab et al 2004b, Keall et al 2004) were
used to optimize the dose delivery for each phase of the breathing cycle. Duan
et al (2003a) showed that the errors introduced by motion did not depend on the
number of fields making up the IMRT treatment. Jiang et al (2003b) also showed
how to incorporate knowledge of the movement of tumour into the actual dynamic
leaf pattern for IMRT.
The gating signal for 4D reconstruction can be obtained by the following
methods:
(i) the phase from an optical respiration monitor,
(ii) the displacement of an external fiducial marker or
(iii) spirometry-based tidal volume.
The main methods to use gated 4D CT data to create maps of individual
voxel movement are:
(i) use of thin plate splines (Rietzel et al 2004, Hartkens et al 2002, Malsch et
al 2004),
(ii) use of optical flow equations (El Naqa et al 2004, Zhang et al 2004a),
(iii) use of viscous fluid flow equations (Mageras et al 2004b) or
(iv) finite element analysis (Brock et al 2004).
Rietzel et al (2004) have made 4D CT scans to show the patient geometry
at 10 different respiratory phases. The Varian RTM system was coupled to a GE
Lightspeed CT Scanner operating in axial cine mode. Using these data, Jiang et al
(2004a) produced a plan that is optimized for each phase (but not overall) when
there is no map of voxel displacements available. Then the set of 10 IMBs for
any fixed gantry angle for the 10 phases can be sequenced by the same algorithm
as used for IMAT because of the analogy between gantry angle and breathing
phase, both being from 0 to 360. This technique is called Synchronized Moving
Aperture Radiation Therapy (SMART) (Jiang 2004). Alternatively, if a voxel
displacement map can be made, e.g. by optical flow mapping (Malsch et al 2004,
El Naqa et al 2004, Zhang et al 2004a, Huang et al 2004, Horn and Schunk
1981), then the objective function summed over all phases can be optimized to
367
produce an IMB set. For any particular phase, the plan is then not optimal.
Deformable alignment has also been done by biomechanical modelling and finite
element analysis (Brock et al 2004) for liver and breast. Zhang et al (2004c) have
created a similar technique to correct for breathing motion during tomotherapy.
Again a voxel displacement map was computed from 4D CT data and the bixel
intensities corresponding to the different breathing phases were computed. Then
the breathing motion was monitored during the Tomotherapy and the appropriate
bixels selected to track the motion.
Kung et al (2003) folded a hypothetical sinusoidal tumour movement into
the leaf patterns of a dMLC treatment for lung cancer and showed acceptable
dose-volume histogram degradations even with one phase of motion. When phase
averaging over fractions was included, the differences were even smaller. The
method relied on the transformation whereby the patient (tumour) was considered
stationary and its movement was instead transferred to the fluence delivery space
of the moving segments.
Chen et al (2001c) have studied the movement of upper thoracic tumours and
their deformation due to breathing and implications for IMRT. Jones and Hoban
(2000) studied the influence of random movement of IMB positions relative to the
patient and showed that a decrease in the EUD ensued.
6.10.12.4 Modelling set-up inaccuracy
Lomax (2001) has studied the effects of inter- and intrafraction positional errors
in radiotherapy for both proton therapy and IMRT. Dose-volume histograms were
computed for specific offsets between planning and image position.
Mock et al (2000) have computed the effect on prostate therapy of
accounting for treatment set-up inaccuracy. Twenty patients were studied and
evaluated using portal images. In all, 186 portal images were analysed. By
feeding the known systematic and random deviations into the HELAX treatmentplanning system and recalculating the dose distribution, it was found that
the TCP averaged over all patients decreased by 19.7% from the anticipated
(no movements included) value of 73.6%. The normal tissue complication
probabilities for bladder and rectum, however, stayed almost unchanged at below
1% and around 5% respectively.
Booth and Zavgorodni (2001) have constructed a Monte Carlo model to
simulate organ translations at the CT imaging stage and to evaluate the effects of
this uncertainty on the dose distribution. The goal was to understand the effect of
a mobile organ not being at its exact mean position at the time of imaging causing
treatment to be planned with an organ offset from its assumed mean position. The
study was limited to 1D movements and the study of single beams.
Baird et al (2001) have studied the effects of set-up uncertainty on dose
distributions for intact breast-treatment techniques. They compared three types
of treatment planning: (1) standard, paired tangential fields with wedges; (2)
electronic compensation and (3) two-field IMRT. Beam isocentres were shifted
368
to represent the average and the maximum set-up uncertainty and the dose was
recalculated for each patient using the optimized fluence patterns. It was found
that both the mean and the maximum patient movement noticeably affected the
dose delivery for the IMRT plans but had no significant impact on the standard
wedged plan or electronic compensation (see also Hector 2001).
Jolly et al (2001) also simulated the effects of organ motion with respect to
tangential breast treatments and concluded that the heterogeneity of dose during
the thoracic movement could lead to a greater probability of tissue complication
due to overdosing the lungs.
Samuelsson et al (2001) have used the University of Michigan treatmentplanning system to show the effects on IMRT of patient-position uncertainty
during the treatment. The degrees of conformality and sharpness of the dose
gradients had a large influence on the sensitivity of IMRT plans to set-up
variations. Inclusion of set-up uncertainties into the optimization process can,
in principle, correct dose distributions that would otherwise be degraded by the
set-up errors.
Manning et al (2001) have investigated the effect of set-up uncertainty on
normal tissue sparing with IMRT for head-and-neck cancer. The first approach
they made was to investigate the effect of adding a planning OAR volume for
the contralateral parotid gland and it was shown that this reduced the radiation
dose to the contralateral parotid gland but slightly worsened the conformality
of dose to the target volume. Then the worse-case scenario was investigated
of systematically translating set-up uncertainties of 5 mm in six orthogonal
directions without altering the optimized beam profiles. It was found that the
plans without the planning OAR volume (PRV) were more susceptible to isocentre
movement than the plans with PRV.
6.10.12.5 Modelling the movement of OARs
Manning et al (2000) have described the concept of PRV in which the effects of
internal motion and set-up margin were added, not only to the PTV but also to the
OAR. It was then shown that optimized nine-beam IMRT plans for three headand-neck patients were unable to achieve their desired outcomes if the motions of
OARs were not taken into account but were able to do so if the PRV was instead
used. This is an interesting study in which the effects of movement are translated
into changes in the dosimetry to OARs.
6.11 Megavoltage CT (MVCT) and kilovoltage CT (kVCT)

for position verification
6.11.1 MVCT
Nakagawa et al (1994) have described how the detector constructed for
megavoltage computed tomography (MVCT) can also be used for real-time beam
Megavoltage CT (MVCT) and kilovoltage CT (kVCT)
369
Figure 6.55. Image of the megavoltage CT system developed in Tokyo. (From Nakagawa
et al 1994)
monitoring in dynamic conformation therapy. The MVCT system developed at

the University of Tokyo in about 1991 (figure 6.55) was a fairly direct copy of
the systems put together by Swindell and colleagues at the University of Arizona
(Swindell et al 1983) and later at the Royal Marsden NHS Foundation Trust,
Sutton. MVCT gives a picture of the patient in the treatment position just prior to
treatment and this takes 38 s using the Tokyo system.
Dynamic conformation is a technique particularly popular in Japan in which,
as the gantry rotates, the leaves of the MLC change their positions to define a new
aperture shape at each 1 of gantry orientation. This does not involve intensity
modulation but is a direct extension of some very early work by Takahashi in the
1950s (see figure 1.1). However, one of the problems with dynamic conformation
therapy is that it is only possible to monitor mispositions in the leaf position with
respect to planning and not with respect to the patient. The idea of Nakagawa et al
(1994) was to use the same 1D detector as is used for MVCT to measure the field
width for the centre pair of the MLC leaves. This was done by measuring the 50%
fluence location to define the field width. This field width was then compared with
that at planning and any mismatch would indicate an error. Deviations between
actual and detected leaf openings have been found to be less than 2 mm for typical
treatment fields. Some disadvantages of this system are that it only measures the
field width for one pair of leaves and, therefore, is not a true 3D verification
370
and, secondly, the quality of megavoltage imaging is not sufficient to delineate

tumour and organ contours. Therefore, the beam positions can only be viewed
with respect to planned beam positions rather than with respect to tumours.
Nakagawa et al (2000a) have shown how MVCT can be used to overcome
treatment inaccuracy during stereotactic radiosurgery for thoracic tumours
(figure 6.56). The MVCT unit comprises a detector array that is mounted on
a linear accelerator at a distance of 160 cm from the beam source. A scan is
performed during a half rotational irradiation taking 35 s. Spatial resolution is
about 3.5 mm at the isocentre level. The slice thickness is 2 cm and the dose
to isocentre is 2.8 cGy per scan. The MVCT is done after patient set-up on the
accelerator couch and used to reposition the patient if necessary. The MVCT
image is then re-used during treatment to superpose the beam, monitored using
a real-time measurement on the detector, on to the MVCT image to show an
accurate direction of the beam at the tumour. Nakagawa et al (2000a) demonstrate
that this improves the reliability of conformal stereotactic radiosurgery and claim
that the use of this MVCT-assisted stereotatic radiosurgery has improved the
clinical outcome for thoracic neoplasms.
Ford et al (2002b) have used an amorphous silicon electronic portal imaging
device to create MVCT images using doses between 10 and 70 cGy. Using just
28 cGy, contrast greater than 5% was distinguishable from background water in
phantom measurements. This is equivalent to the use of just 35 MU.
Pouliot et al (2002) and Ghelmansarai et al (2003) have made similar studies
and shown that MVCT is possible with just 15 MU showing visualization of bony
anatomy structures adequate for patient positioning. Pouliot et al (2003) then
argued that such MVCT images can be correlated to planning CT images to create
the required relocation parameters for each fraction.
Guan (2002) has performed MVCT using 50 full projections and 50
truncated projections each with 1 MU and claims that, with an amorphous silicon
500 imaging device, it is possible to produce 1.5% contrast detectability. Nearly
double the number of MUs is required for a CCD portal-imaging system also
performing MVCT.
Hajdok et al (2002) have made MVCT images using a 60 Co source. They
claim 3 mm high-contrast spatial resolution and 2.8% low-contrast sensitivity,
linearity between MVCT number and electron density and an inherent absence of
beam-hardening artefacts.
Partridge and Hesse (2002) have constructed MVCT images at the time of
treatment based on the use of a portal-imaging device and then backprojected
the primary fluence from the measured EPID signal through the MVCT image to
construct the input fluence maps for IMRT which may be directly compared with
the predicted fluence maps (see section 3.13.1).
Munro et al (2002) and Seppi et al (2003) have developed yet another piece
of apparatus for MVCT (figure 6.57). This is a flat-panel imager comprising a
scintillator of individual CsI crystals 8 mm thick by 0.38 mm 0.38 mm pitch
with five sides of each crystal coated with a reflecting powder/epoxy mixture and
371
Figure 6.56. The figure shows a series of displays of a megavoltage computed tomography
image of the thorax of a patient. Superimposed on this image are displays of the beam
direction and shape displaying in real-time monitoring of the MLC for the rotational
conformation technique. (Reprinted from Nakagawa et al (2000a) with copyright
the sixth side in contact with the flat-panel sensor. The flat-panel imaging system
enables projections to be taken in just two radiation pulses which corresponds
to 0.046 cGy dose. By collecting 360 projections, an MVCT image could be
created by reconstruction using just 16 cGy of dose and reconstructing with the
Feldkamp conebeam CT reconstruction algorithm. Experiments showed that the
contrast resolution was about 1% for large objects and high-contrast structures
such as air holes and embedded seeds could be seen with a spatial size of about
1.22.4 mm in size (figure 6.58). Images can be made with 1 MU per projection.
Images with just a total dose of 16 cGy show soft-tissue structure such as the
heart, lung, kidneys and liver in reconstructed images of an anthropomorphic
phantom. This system bears great resemblance to that constructed by Mosleh
Shirazi et al (1998). The authors commented on the lengthy 20-year history of
the development of MVCT and they also suggest that, if their detector could be
improved, it would be possible to image soft tissue such as the prostate. At the
moment, this is not possible and also they comment that 16 cGy dose is probably
372
still too large to be used on a daily basis if one is imaging the entire patient cross
section. Compromises might include the performance of local tomography.
Mageras et al (2004a) made MVCT images with the Varian 4030 HE
EPID in cone beam geometry obtaining images with a density resolution of 2%.
This detector has an 8-mm-thick pixellated CsI scintillation layer on top of the
amorphous silicon photodiodes and a detective quantum efficiency (DQE) some
10 times that of copper-plus-Gadox systems. By gating with the Varian RTM
imaging system, 4D CT is also possible.
6.11.2 Flat-panel imaging for kVCT
Jaffray (2002, 2003) and Jaffray et al (2002) have developed an imaging system
that generates high-resolution soft-tissue images of the patient at the time of
treatment with the purpose of guiding radiotherapy and reducing uncertainties
(figures 6.59 and 6.60). It uses a flat-panel x-ray detector mounted opposite
a kV tube on a retractable arm at 90 to the treatment source. A filteredbackprojection algorithm is used to perform cone-beam reconstructions and yield
images 0.1 cm thick with an imaging exposure of 1.2 ro ntgens and giving a
contrast of 47 Hounsfield units. These images, in principle, enable patient
positioning at the time of treatment to be adjusted to correct for errors of
alignment.
This was the basis of the Elekta iViewGT release 2 which supported IMRT
via step-and-shoot. This is a 41 cm 41 cm amorphous silicon detector projecting
back to 26.2 cm 26.2 cm at isocentre and with a 15 cm offset. It has a motorised
retractable arm. Release 3 supports the flat-panel imager with a retractable arm
and very good image quality. Image registration using the William Beaumont
template mapping is possible. Chamfer matching with The Netherlands Cancer
Institute software and seed recognition with the University of Utrecht software
(see also Wavelength 2001d) is possible and it also supports cone-beam CT (see
Wavelength 2003a).
The commercial product from Elekta is called the Elekta Synergy system
(Wavelength 2003e, 2004) and has been developed and evaluated through a
consortium of four centres:
(i)
(ii)
(iii)
(iv)
William Beaumont Hospital, Royal Oak;

The Netherlands Cancer Institute, Amsterdam (van Herk et al 2002b);
Princess Margaret Hospital, Toronto;
Christie Hospital, Manchester (figure 6.61) (Williams 2003a).
Van Herk et al (2002b) have used kVCT on a treatment machine to localize

the prostate prior to radiotherapy. To do this, they developed automatic contourextraction methods and registration techniques so that the patient could be moved
to the planning position. Artignan et al (2002a) give more details. Van Herk et al
(2004) used kVCT images to obtain interfraction shifts to realign the patient.
373
Figure 6.57. The experimental set-up for megavoltage computed tomography. The
phantom was rotated with the x-ray source and the image receptor stationary. A cylindrical
normalization (polyethylene) phantom is shown mounted on the rotating stage. (Reprinted
from Seppi et al (2003) with copyright permission from Elsevier.)
374
Figure 6.58. Axial, coronal and sagittal CT slices (5 mm thick) of the head phantom
acquired at 6 MV using a 16 cGy irradiation. (Reprinted from Seppi et al (2003) with
Zijp et al (2004) showed how to create the Amsterdam Shroud from the
kVCT cone-beam images which can be image processed to yield the diaphragm
motion without any markers and so obtain the image phase respiratory signal.
Letourneau et al (2002) have shown that the Elekta/Jaffray kilovoltage
portal-imaging system can be used for setup verification for radiotherapy.
Nusslin et al (2003) have introduced image-guided radiotherapy into their
HYPERION treatment-planning system. They have tools to adjust the contours
within the planning system and to evaluate the effects of such adjustments on
treatment outcome. These adjustments might be signalled if it were possible to
re-image the patient between each treatment fraction. The use of an amorphous
silicon flat-panel imager could assist such fraction-by-fraction adjustments.
6.12 MRI and IMRT simultaneously

Lagendijk and Bakker (2000), Langendijk et al (2002) and Raaymakers et al
(2003) have developed the concept of a low-field intensity MRI unit integrated
with a low-energy accelerator both with the same isocentre. The MRI unit is a
0.20.3 T machine. This enables MRI images to be taken of the patient in the
treatment position and used as the basis for treatment planning and also MRI
images to provide the basis for on-line position verification which is claimed to
be far superior to megavoltage position verification. Raaymakers et al (2004)
have completed the design study. Active magnetic shielding will uncouple the
MRI system and the accelerator. Modifications to the coil windings reduce
the attenuating path of the rays from the external accelerator to 10 cm of
MRI and IMRT simultaneously
375
Figure 6.59. (a) A medical linear accelerator has been modified for kV cone-beam
computed tomography. A kV x-ray tube has been mounted on a retractable arm at
90 with respect to the treatment source. A large-area (41 cm 41 cm) flat-panel
imager is mounted opposite the kV x-ray tube on a fixed mount. (b) A photograph of
the large-area flat-panel detector employed in this investigation (black) in comparison
to a smaller detector employed in previous investigations and an anthropomorphic head
phantom. Labels indicate the pixel format and the pixel pitch of the imagers as well as
the approximate sensitive area. The pitch is quoted in micrometres and the pixels are the
number of pixels. (Reprinted from Jaffray et al (2002) with copyright permission from
Elsevier.)
376
Figure 6.60. A small Hitchcockian cameo as the author props himself up next to the
prototype kVCT machine at the William Beaumont Hospital, Royal Oak in October 2003.
(Photograph by Dr Mark Oldham.)
aluminium. The prototype is being considered by a manufacturer but is not yet

under construction.
6.13 IMRT using mixed photons and electrons

Svatos et al (2000) have demonstrated the advantages of using mixed beams of
21 MeV electrons and several configurations of intensity-modulated 6 MV photon
beams. The addition of a single electron beam to the target reduced the integral
dose two-fold.
Dogan et al (2001a) have studied how to match IMRT photon plans with
electron beams using the CMS FOCUS treatment-planning system. It was found
that, in order to avoid the production of either hot spots or cold spots, a very
precise matching of the fields was required and the precise gap between the fields
was specifically important. For the case studied, the idealized gap was 3 mm.
Dogan et al (2002a) noted that when IMRT photon fields and electron
fields are matched, hot spots of up to 145% arose due to the bulging of the
electron penumbra. They designed techniques to improve the field matching.
These required the PTV to be extended into the region previously covered by
IMRT using mixed photons and electrons
377
Figure 6.61. A view of the Christie Hospitals research kVCT machine showing the kV
source mounted at 90 to the treatment head. The kV source is shown in the extended
position and in line with the amorphous silicon flat panel. This is the system of Jaffray et
al (2002) commercialized by Elekta Oncology Systems. (From Wavelength 2002.)
the electrons. The high isodose lines were then designed to bend away from the
bulging electron penumbra. The electron penumbra was also further modified
by using an MLC. Experimental measurements were performed abutting both
tomographic IMRT delivery, step-and-shoot delivery (Varian accelerator) with
the electron fields. The film calibration was performed using mixed photon and
electron irradiation to properly account for the different spectral effects. It was
reported that, with no error in the field matching, the hot spots dropped to 105%
in a test case and were still at only 115% when a 1.5 mm matchline error was
deliberately introduced.
Korevaar et al (2002) have made comparative planning investigations to
decide whether the additional use of high-energy electrons in photon IMRT can
improve on predicted outcome. Studies showed that largely this was not the
case. IMRT plans with and without electrons were very similar and it was
recommended that high-energy electrons should not be used.
Ma et al (2003c) have compared tangential photon breast radiotherapy,
IMRT and modulated electron radiotherapy (MERT) for breast cancer treatment.
It was concluded that photon IMRT provided superior target homogeneity
378
compared with conventional tangents but that MERT offered significant

improvement over tangents for the plans inspected. The breathing margin could
be significantly reduced because electron beams do not require direct tangential
fields.
Xing et al (2003) have developed a technique to generate uniform doses from
matched photon and electron fields. First, the electron plan was created and then
the IMRT photon plan was created to dovetail to it along a matchline.
Epilogue
Consider IMRT as a train and the development of IMRT as like a long train
journey. The train started in a cold siding a long way from its destination. This
siding is where radiotherapy was in, say, the start of the 1980s. The destination
is a station at which IMRT is at a state of development whereby it can be wisely
but widely applied to all patients who need it, with quality assured and with due
notice taken of all the potential impediments. The professionals developing IMRT
certainly would like the train to be at this destination right here and now. So would
the patients. They need to have a technology providing either a cure or, if this is
unachievable, a high quality for their remaining life. Those who manage our
health carehospital managers, funders, fundraisers and politiciansalso would
like the train now in this station. However, the train is still some way from the
destination.
En route, the IMRT train has stopped in several stations with mileposts. The
first had the basic concepts of planning worked out (about 1990). The second
had the fundamental concepts of IMRT delivery established but only in a few
research centres (1994). The train got an extra locomotive in the late 1990s as
companies began to market the needed equipment. So the train went faster and,
by 2000, IMRT was a technology whose constituent elements one could largely
purchase rather than create in-house (company-sponsored courses aided education
and provided some interesting IMRT souvenirs [figure E.1]). At other stations, 3D
medical imaging was coupled to the train. Also en route at stations the coaches
which will carry the passengers have been upgraded. Some new coaches, like
spiral tomotherapy, which werent there at the start of the journey have been
hooked up. Quality assurance, that at the start of the journey needed a person
to tap the wheels of every coach before proceeding (time consuming), has been
speeded up and now fewer checks need to be made per stop. However, from time
to time at stations the train itself needs a larger inspection and overhaul. In short,
the train that it is hoped will arrive at the destination is almost unrecognizably
different from the one at the start of the journey. The trainspotting manuals (this
series of books) have evolved to describe these landmarks.
The railway officials are the doctors, physicists, radiographers, engineers, all
those professionals whose skills combine to move the IMRT train safely. Like
railway officials of old, they work together, the only sensible way to ensure the
380
Epilogue
Everyone will
receive one of these
from MRC-Systems
+ a certificate on
Saturday at the end
of the course.
Figure E.1. Two examples of IMRT souvenirs produced by companies sponsoring IMRT
teaching courses. The mug on the left was given to all the students at the MRC-DKFZ
Heidelberg Schools. The hat on the right was worn by delegates at the first-ever IMRT
course in Durango, CO in May 1996.
passage of the train. We should avoid any possibility that those who drive the
trains separate from those who run the track. Research and development need to
go hand in hand with clinical service. No one professional group owns the IMRT
railway. Successful ongoing progress depends on the collaboration.
But there are leaves on the line (the UK readers will appreciate the analogy)
and the safety of the train has been questioned. We are quite adept at delivering
highly conformal dose distributions to immobile phantoms and also to some
patients in certain well-controlled situations. But patient movement makes up
some of the leaves on the line. So does our growing awareness that, whereas once
diagnostic identification of disease was superior to the ability to treat it, maybe
now the reverse is true. The functional status of tissues needs to be established.
We would like to know the distribution of clonogenic cells. We would like to
establish the relationship between the disease which shows as changed x-ray
attenuation and that which shows as changed function. These are more leaves
on the line. But the train must not stop. Instead, with wet leaves, it must proceed
cautiously whilst ways to sweep away the leaves or ride safely over them are
considered.
Epilogue
381
Analogies pushed too far usually begin to falter and some will have seen
some flaws in this one (since the journey is through time, passengers (patients)
have been joining this train at different states of its development). So enough
of trains. However, the message is clear. IMRT has come a long way from
the days, not much more that ten years ago when the first patient was yet to be
treated and IMRT was (as I have heard it unfortunately described) a physicists
toy. The challenge is now to solve those final problems, then to roll out the
technology as widely as possible (a) so as many benefit as require it and (b)
to establish objectively through phase-3 randomized trials the benefit of IMRT.
There will always be arguments of costbenefit in a society with competing needs
for so many different forms of healthcare. I believe IMRT will meet a clear need,
lead to objectively assessable outcomes and justify the optimism of its developers
and supporters. As I am no prophet and sadly also do have a pessimistic side, I
expect it to throw up new questions and reach forms that we have yet not begun
to imagine.
References
Aaronson R, DeMarco J, Chetty I and Solberg T 2001 A Monte Carlo model for quality
assurance of intensity modulated radiotherapy incorporating specific leaf constraints
Med. Phys. 28 1268
Achterberg N and Mueller R G 2001 Multifocal static tomotherapy with multileaf
collimation and table movement Radiother. Oncol. 61 S58
Adams E J, Convery D J, Cosgrove V P, McNair H A, Staffurth J N, Vaarkamp J,
Nutting C M, Warrington A P, Webb S, Balyckyi J and Dearnaley D P 2004 Clinical
implementation of dynamic and step-and-shoot IMRT to treat prostate cancer with
high risk of pelvic lymph node involvement Radiother. Oncol. 70 110
Adams E J, Convery D J, Warrington A P and Webb S 2001 IMRT planning at the Royal
Marsden NHS Trust using TMS Insights 3 12
Adams E, Vaarkamp J, Convery D J, Partridge M, Warrington A P and Webb S 2003
Refinement of a quality assurance programme for IMRT Clinical Oncol. 15 (Suppl.
2) S7
Adams E J, Vaarkamp J, Convery D J, Warrington A P, McNair H and Webb S 2002 Quality
assurance for IMRT: a diminishing burden Radiother. Oncol. 64 (Suppl. 1) S125
Agazaryan N, Llacer J, Ullrich W, Promberger C, Solberg T, Arellano A and Paul T 2001
Deliverability scoring and dosimetric investigation of leaf sequencing in IMRT Med.
Phys. 28 1253
Agazaryan N and Solberg T D 2003 Segmental and dynamic intensity modulated
radiotherapy delivery techniques for micro multileaf collimator Med. Phys. 30 1758
67
Agazaryan N, Solberg T, Arellano A R and Paul T 1999 Leaf sequencing for fluence
modulated radiation therapy Med. Phys. 26 1139
2000 Three-dimensional verification for dynamic multileaf collimated IMRT Proc.
World Congress of Medical Physics and the AAPM Annual Congress, August 2000
paper MO-BBR-02
Agazaryan N, Solberg T and Llacer J 2002 An investigation of deliverability complexity
in intensity modulated radiotherapy Med. Phys. 29 1197
Alaei P, Higgins P, Weaver R and Nguyen N 2002 Comparison of dynamic and step-andshoot intensity-modulated radiation therapy planning and delivery using the Helios
system Med. Phys. 29 1270
Alber M 2001a Enhanced biological planning and expedited delivery of IMRT Radiother.
Oncol. 61 (Suppl. 1) S75
2001b Optimisation of IMRT under contraints for static and dynamic MLC delivery
Med. Phys. 28 1205-6
References
383
Alber M, Birkner M, Laub W and Nusslin F 2000 Optimisation of IMRT under biological
and practical constraints using Monte Carlo dose computation Radiother. Oncol. 56
(Suppl. 1) S79
Alber M, Fippel M and Nusslin F 2001a IMRT optimisation with Monte Carlo dose
computation including the MLC Radiother. Oncol. 61 (Suppl. 1) S64
Alber M, Meedt G, Birkner M and Nusslin F 2002b Universal properties of the IMRT
optimisation problem and consequences for the optimum number of beams and MLC
field segments: conflict spectroscopy Radiother. Oncol. 64 (Suppl. 1) S96
Alber M, Meedt G, Nusslin F and Reemtsen R 2002a On the degeneracy of the IMRT
optimisation problem Med. Phys. 29 25849
Alber M and Nusslin F 2001 Optimization of intensity modulated radiotherapy under
constraints for static and dynamic MLC delivery Phys. Med. Biol. 46 322939
Aletti P, Marchesi V, Bey P, Lapeyre M, Beckendorf V, Marchal C and Noel A 2002
Clinical implementation of the IMRT at the A. Vautrin Hospital (NancyFrance)
Radiother. Oncol. 64 (Suppl. 1) S216
Aletti P, Metayer Y, Bey P, Beckendorf V, Buchheit I, Noel A and Marchal C 2001
Comparison of IMRT and standard conformal therapy for the treatment of the cancer
of the prostate Radiother. Oncol. (Suppl. 1) 61 S67
Al-Ghazi M, Dunia R, Sanford R and Radany E 2002 When is a combination of
IMRT/conformal radiation therapy superior to either IMRT or conformal therapy
alone? Med. Phys. 29 1214
Alheit H, Dornfeld S, Winkler C, Blank H and Geyer P 2000 Stereotactic guided irradiation
in prostatic cancer using the ExacTrac-System (BrainLab) Radiother. Oncol. 56
(Suppl. 1) S107
Allen A M, Siracuse K M, Hayman J A and Balter J M 2004 Evaluation of the influence of
breathing on the movement and modeling of lung tumors Int. J. Radiat. Oncol. Biol.
Phys. 58 12517
Amols H 2003 Snake oil: Whos buying and whos selling? in regard to Glatstein: The
return of the snake oil salesmen Int. J. Radiat. Oncol. Biol. Phys. 56 1507
2004 In response to Drs. Nutting and Harrington (re: Glatstein/Amols, Snake Oil) Int.
J. Radiat. Oncol. Biol. Phys. 58 1317
Andre L and Haas B 2000 A comparison between intensity modulated treatment fields
generated with sliding window algorithm and with the step and shoot algorithm Proc.
paper WE-E313-05
Aoki Y and Yoda K 2003 A multi-portal compensator system for IMRT delivery Med.
Phys. 30 1347-8
Aoyama H, Shirato H, Nishioka T, Hashimoto S, Tsuchiya K, Kagel K, Onimaru
R, Watanabe Y and Miyasaka K 2001 Magnetic resonance imaging system for
three-dimensional conformal radiotherapy and its impact on gross tumor volume
delineation of central nervous system tumors Int. J. Radiat. Oncol. Biol. Phys. 50
8217
Arellano A, Solberg T and Llacer J 2000 Clinically oriented inverse planning
implementation Proc. World Congress of Medical Physics and the AAPM Annual
Congress, August 2000 paper WE-E313-02
Arnfield M R, Siebers J V, Kim J O, Wu Q, Keall P J and Mohan R 2000 A method
for determining multileaf collimator transmission and scatter for dynamic intensity
modulated radiotherapy Med. Phys. 27 223140
384
References
Artignan X, Smitsmans M H P, de Bois J, Jaffray D, Ghilezan M, Fournerel P, Steenbakkers

R, Lebesque J V, van Herk M and Bartelink H 2002a On-line image guided
radiotherapy for prostate cancer: an accelerated method for prostate localization
Artignan X, Smitsmans M H P, de Bois J, Lebesque J V, van Herk M and Bartelink H
2002b On-line ultrasound-image guidance for radiotherapy of prostate cancer: the
impact of image acquisition on prostate displacement Radiother. Oncol. 64 (Suppl.
1) S279
Artignan X, Smitsmans M H P, Lebesque J V, Jaffray D A, van Herk M and Bartelink H
2004 Online ultrasound image guidance for radiotherapy of prostate cancer: impact
of image acquisition on prostate displacement Int. J. Radiat. Oncol. Biol. Phys. 59
595601
Aspradakis M M, Walker C P, Steele A and Lambert G D 2002 Accelerator commissioning
for IMRT: small MU segments and small fields Radiother. Oncol. 64 (Suppl. 1) S62
Ayyangar K M, Li S, Saw C B, Yoe-Sein M, Pillai S, Chivukula V S and Enke C 2003
Comparison of a complex prostate plan using three IMRT planning systems Med.
Phys. 30 1487
Azcona J D, Siochi R A and Azinovic I 2002 Quality assurance in IMRT: Importance of
the transmission through the jaws for an accurate calculation of absolute doses and
relative distributions Med.Phys. 29 26974
Aznar M C, Pignol J P, Sixel K E and Benk V A 2001 Improving dose homogeneity in
breast irradiation: a dose volume histogram study Med. Phys. 28 1988
Aznar M C, Sixel K E and Ung Y C 2000 Feasibility of deep inspiration breath hold
combined with intensity modulated radiation treatment delivery for left sided breast
cancer Proc. 42nd Annual ASTRO Meeting. Int. J. Radiat. Oncol. Biol. Phys. 48
(Suppl.) 297
Backfrieder W, Fransson-Andreo A, Lorang T, Nowotny R, Dieckmann K and Poetter R
2000 Tissue classification of MR images for use in radiotherapy planning Radiother.
BackSamuelsson A and Johansson K A 2002 Head scatter on and off axis in small MLC
defined beams applied for IMRT treatments with dynamic MLC Radiother. Oncol.
64 (Suppl. 1) S37
Baird C, Turian J, Myrianthopoulos L and Hamilton R 2001 Effects of setup uncertainty
on dose distribution from intact breast treatment techniques Med. Phys. 28 1292
Bakai A, Alber M and Nusslin F 2002 Fast quality assurance for highly modulated IMRT
beams employing multiple point measurement acceptance criteria Radiother. Oncol.
64 (Suppl. 1) S211
Baker C, Wolff T and Fenwick J 2002 Correction of monitor units for head scatter
contributions in asymmetric IMRT fields from a commercial treatment planning
system Radiother. Oncol. 64 (Suppl. 1) S36
Baldock C, Murry P and Kron T 1999 Uncertainty analysis in polymer gel dosimetry Phys.
Med. Biol. 44 N2436
Balog J B, Caldwell C B, Ung Y C, Mah K, Danjoux C E, Ganguli S N and Ehrlich L
E 2000 Interobserver variation in contouring gross tumour volume in carcinoma of
the lung associated with pneumonitis and atelectasis: The impact of 18FDG-hybrid
PET fusion Proc. 42nd Annual ASTRO Meeting. Int. J. Radiat. Oncol. Biol. Phys. 48
(Suppl.) 128
References
385
Balog J, Olivera G and Kapatoes J 2003 Clinical helical tomotherapy commissioning

dosimetry Med. Phys. 30 3097106
Balter J 2003 Demonstration of accurate localization and continuous tracking of
implantable wireless electromagnetic transponders Med. Phys. 30 1382
Balter J, Brock K, Kashani R, Coselmon M, Kessler M and McShan D 2003a The evolution
of a dynamic lung model for treatment planning Radiother. Oncol. 68 (Suppl. 1) S41
Balter J M, Litzenberg D W, Brock K K, Sanda M, Sullivan M, Sandler H M and Dawson L
A 2000 Ventilatory movement of the prostate during radiotherapy Proc. 42nd Annual
ASTRO Meeting. Int. J. Radiat. Oncol. Biol. Phys. 48 (Suppl.) 167
Balter J, Wright N, Dimmer S, Friemel B, Newell J, Cheng Y and Mate T 2003b
Demonstration of accurate localisation and continuous tracking of implantable wires
Int J. Radiat. Oncol. Biol. Phys. 57 S264
Balycyki J, Cattell A, Bloomfield D and Powley S 2001 Commissioning radiotherapy
equipmenta multidisciplinary approach RAD Magazine 27 (233) 213
Bamber J C, Verwey J A A, Eckersley R J, Hill C R and ter Haar G 1996 Potential for tissue
movement compensation in conformal cancer therapy Acoustical imaging vol 22, ed
P Tortoli and L Masotti (New York: Plenum) pp 23944
Bar W, Alber M, Birkner M, Mondry A, Bakai A and Nusslin F 2001a Implementation of a
step and shoot IMRT concept under consideration of clinical and dosimetric aspects
Med. Phys. 28 1202
Bar W, Alber M and Nusslin 2001b A variable fluence step clustering and segmentation
algorithm for step and shoot IMRT Phys. Med. Biol. 46 19972007
2001c A method to reduce the number of beam segments for step and shoot IMRT
Bar W, Schwarz M, Alber M, Bos L J, Mijnheer B J, Rasch C, Schneider C, Nusslin F
and Damen E M F 2003 A comparison of forward and inverse treatment planning
for intensity-modulated radiotherapy of head and neck cancer Radiother. Oncol. 69
2518
Bar W, Schwarz M, Rasch C, Alber M, Schneider C, Bos L, Mijnheer B, Lebesque
J, Damen E and Nusslin F 2002 A treatment planning comparison for intensitymodulated radiotherapy of head and neck cancer Radiother. Oncol. 64 (Suppl. 1)
S124
Barentsz J 2000 Prospects of contrast enhanced MR imaging for radiotherapy in the pelvis
Barker J, Patrocinio H and Podgorsak E B 2001 A comparison study of multileaf and
micro-multileaf collimators Med. Phys. 28 1986
Barnes E A, Murray B R, Robinson D M, Underwood L J, Hanson J and Roa W H Y
2001 Dosimetric evaluation of lung tumour immobilization using breath hold at deep
inspiration Int. J. Radiat. Oncol. Biol. Phys. 50 10918
Baroni G, Ferrigno G, Orecchia R and Pedotti A 2000 Real-time opto-electronic
verification of patient position in breast cancer radiotherapy Computer Aided Surgery
5 296306
Barthold S, Echner G, Hartmann G, Pastyr O and Schlegel W 2002 CoRAA new cobalt
radiotherapy arrangement with multiple sourcesa feasibility study Phys. Medica.
19 1326
Bassalow R and Sidhu N 2003 Dark current radiation produced in step-and-shoot IMRT
on clinac 21EX linear accelerator Med. Phys. 30 1493
386
References
Bate T, de Meerleer G, Bakker M, Villeirs G and de Neve W 2002 IMRT for prostate
cancer: the consequences on the workload Radiother. Oncol. 64 (Suppl. 1) S889
Bate M T, Vakaet L and de Neve W 2000 Quality control and error detection in the
treatment process Radiother. Oncol. 56 (Suppl. 1) S89
Battista J and Baumann J S 2003 The future of IMRT IMRTthe State of the Art: AAPM
Medical Physics Monograph Number 29 ed J R Palta and T R Mackie (Madison, WI:
Medical Physics Publishing) pp 87582
Baumert B G, Ciernik I, Dizendorf E, Reiner B, Davis J B, VonSchulthess G K, Lutolf U
M and Steinert H C 2002 Implication of combined PET-CT on decision making in
radiation oncology Radiother. Oncol. 64 (Suppl. 1) S209
Bayouth J, Pena J, Culp L, Brack C and Sanguineti G 2003a Feasibility of IMRT to cover
pelvic nodes while escalating the dose to the prostate gland: dosimetric and acute
toxicity data on 24 consecutive patients Radiother. Oncol. 68 (Suppl. 1) S77
Bayouth J E and Morrill S M 2003c MLC dosimetric characteristics for small field and
IMRT applications Med. Phys. 30 254552
Bayouth J E, Wendt D and Morrill S M 2003b MLC quality assurance techniques for IMRT
applications Med. Phys. 30 74350
Beaulieu F, Beaulieu L, Tremblay D, Lechance B and Roy R 2004 Automatic generation
of anatomy-based MLC fields in aperture-based IMRT Med. Phys. 31 153945
Beavis A 2003a Dynamic contrast enhanced MRI in the inverse planning of IMRT: Prostate
case study Clinical Oncol. 15 (Suppl. 2) S56
2003b IMRT in our real NHS world? RAD Magazine 29 (340) 334
2004a The need for more intelligence in radiotherapy RAD Magazine 30 (346) 312
2004b Is tomotherapy the future of IMRT? Brit. J. Radiol. 77 28595
Beavis A, Bubula E and Whitton V 2002 Pre-treatment quality assurance for IMRT: the
Princess Royal Hospital protocol Radiother. Oncol. 64 (Suppl. 1) S85
Beavis A, Ganney P, Whitton V and Xing L 2000a Optimisation of MLC orientation
to improve accuracy in the static field delivery of IMRT Proc. World Congress of
Medical Physics and the AAPM Annual Congress, August 2000 paper FR-B309-02
2000b Slide and shoot: a new method for MLC delivery of IMRT. Proc. 13th Int. Conf.
on the Use of Computers in Radiation Therapy (Heidelberg, May) ed W Schlegel and
T Bortfeld (Heidelberg: Springer) pp 1824
2001 Optimization of the step-and-shoot leaf sequence for delivery of intensity
modulated radiation therapy using a variable division scheme Phys. Med. Biol. 46
245765
Beavis A W, Garcia-Alvarez R and Liney G P 2003 The use of FMRI in IMRT treatment
planning: A case study Med. Phys. 30 13845
Beavis A and Liney G 2002a Delineation of IMRT boost volumes for prostate treatments
using advanced magnetic resonance (DCE) techniques Med. Phys. 29 1257
2002b The application of an advanced magnetic resonance technique for planning
boost delivery with IMRT Radiother. Oncol. 64 (Suppl. 1) S21011
Beavis A W, Whitton V I and Xing L 1999 A combination delivery mode for intensity
modulated radiation therapy based on specific patient cases. Int. J. Radiat. Oncol.
Biol. Phys. 45 (Suppl. 1) 164
Bedford J L 2003 Optimisation and inverse planning for intensity modulated radiotherapy:
Fundamental concepts and practical factors Recent Res. Dev. Radiol. 1 2338
References
387
Bedford J L and Webb S 2003 Elimination of importance factors for clinically accurate
selection of beam orientations, beam weights and wedge angles in conformal
radiation therapy Med. Phys. 30 1788804
Bednarz G, Michalski D, Houser C, Huq M S, Xiao Y, Anne P R and Galvin J M 2002 The
use of mixed-integer programming for inverse treatment planning with pre-defined
segments Phys. Med. Biol. 47 223545
Beitler J J 2003 Lets accelerate the PET learning curve Int. J. Radiat. Oncol. Biol. Phys.
55 281
Bel A, Astreinidoi E, Dehnad H and Raaijmaakers C P J 2002 Evaluation of the effect
of geometric uncertainties on IMRT plans for head and neck treatments Radiother.
Benard F, Boucher L, Boisvert C, Boileau R, Laberge G and Nabid A 2000 The use of
FDG-PET imaging for the management of lung cancer patients Radiother. Oncol. 56
(Suppl. 1) S40
Benedict S H, Cardinale R M, Wu Q, Zwicker R D, Broaddus W C and Mohan R
2001 Intensity-Modulated stereotactic radiosurgery using dynamic micro-multileaf
collimation Int. J. Radiat. Oncol. Biol. Phys. 50 7518
Bentzen S M 2004 High-tech radiation oncology: should there be a ceiling? Int. J. Radiat.
Oncol. Biol. Phys. 58 32030
Berbeco R I, Mostafavi H, Sharp G C and Jiang S B 2004 Tumor tracking in the absence
of radiopaque markers Proc. 14th Int. Conf. on the Use of Computers in Radiation
Therapy (Seoul, May) pp 4336
Beyer D, Quiet C, Rogers K and Gurgoze E 2000 Prostate motion during a course of IMRT:
the value of daily ultrasound to guide portal localization and correct for random and
systematic errors Radiother. Oncol. 56 (Suppl. 1) S105
Bieda M, Szeglin S and Das I 2002 Microanalysis and temporal characterization of small
MU segments in IMRT Med. Phys. 29 1304
Bignardi M, Galelli M, Spiazzi L, Bonetti M and Magrini S M 2000 Conformal
radiotherapy for prostate carcinoma: differences in dose distribution between the
original plan and the same plan recalculated (in the same patient) after 3 and 6 weeks
of treatment Radiother. Oncol. 56 (Suppl. 1) S108
Birkner M, Alber M, Yan D and Nusslin 2002 Image guided adaptive IMRT of the prostate
based on a probabilistic patient geometry Radiother. Oncol. 64 (Suppl. 1) S282
Birkner M, Yan D, Alber M, Jiang L and Nusslin F 2003 Adapting inverse planning to
patient and organ geometrical variation: algorithm and implementation Med. Phys.
30 282231
Birkner M, Yan D and Nusslin F 2001 Incoporating organ motion and patient setup into
inverse planning using patient specific image feedback Radiother. Oncol. 61 (Suppl.
1) S13
Blackall J, Landau D, Ahmad S, Crum W, McLeish K and Hawkes D 2003 Image
registration based modelling of respiratory motion for optimisation of lung cancer
radiotherapy Proc. World Congress in Medical Physics (Sydney) CDROM and
website only
Bliss P and Wilks R J 2002 The use of a compensator library to reduce dose inhomogeneity
in tangential radiotherapy of the breast Radiother. Oncol. 64 (Suppl. 1) S134
Boehmer D, Bohsung J, Eichwurzel I, Moys A and Budach V 2004 Clinical and physical
quality assurance for intensity modulated radiotherapy of prostate cancer Radiother.
Oncol. 71 31925
388
References
Bohsung J, Groll J, Kurth C, Pfaender M, Grebe G, Jahn U, Wurm R E, Stuschke M

and Budach V 2000 Clinical implementation of the VARIAN intensity modulated
radiotherapy (IMRT) system Radiother. Oncol. 56 (Suppl. 1) S195
Boman E, Lyyra-Laitinen T, Kolmonen P, Jaatinen K and Tervo J 2003 Simulations for
inverse radiation therapy treatment planning using a dynamic MLC algorithm Phys.
Med. Biol. 48 92542
Bookstein F 1989 Principal Warps: Thin plate splines and the decomposition of
deformations IEE Trans. Patt. Anal. Mach. Intell. 11 56785
Booth J T and Zavgorodni S F 2001 Modelling the dosimetric consequences of organ
motion at CT imaging on radiotherapy treatment planning Phys. Med. Biol. 46 1369
77
Borrosch D, Micke O, Papke K, Schaefer U, Schuck A, Schueller P and Willich N 2000
Functional MRI in the treatment planning of conformal radiotherapy for reirradiation
of malignant brain tumors Proc. 42nd Annual ASTRO Meeting. Int. J. Radiat. Oncol.
Biol. Phys. 48 (Suppl.) 258 (Handout material of the AAPM Meeting in 1999)
Bortfeld T 2000 Step and shoot IMRT Radiother. Oncol. 56 (Suppl. 1) S83
2001 Current status of IMRT: physical and technological aspects Radiother. Oncol.
61 (Suppl. 1) S11
2003 Physical optimisation IMRTthe State of the Art: AAPM Medical Physics
Monograph Number 29 ed J R Palta and T R Mackie (Madison, WI: Medical Physics
Publishing) pp 5175
Bortfeld T R, Boyer A L, Schlegel W, Kahler D L and Waldron T J 1994 Realisation and
verification of three-dimensional conformal radiotherapy with modulated fields Int.
Bortfeld T, Burkelbach J, Boesecke R and Schlegel W 1990 Methods of image
reconstruction from projections applied to conformation therapy Phys. Med. Biol.
35 142334
Bortfeld T, Jiang S B and Rietzel E 2004 Effects of motion on the total dose distribution
Seminars in Radiat. Oncol. 14 4151
Bortfeld T, Jokivarsi K, Goitein M and Jiang S 2002b Effects of intra-fraction motion on
IMRT delivery with MLC, compensators and scanning Med. Phys. 29 1347
Bortfeld T, Jokivarsi K, Goitein M, Kung J and Jiang S B 2002a Effects of intra-fraction
motion on IMRT dose delivery: statistical analysis and simulation Phys. Med. Biol.
47 220320
Bortfeld T, Kufer K H, Monz M, Scherrer A, Thieke C and Trinkaus H 2002c Radiation
therapy planninga multicriteria approach Med. Phys. 29 1258
Bortfeld T, Oelfke U and Nill S 2000 What is the optimum leaf width of a multileaf
collimator? Med. Phys. 27 2494502
Bortfeld T, Schlegel W, Hover K-H and Schulz-Ertner D 1999 Mini and micro multileaf
collimatorshandout material; 41st AAPM meeting, Nashville, TN, July 1999
Bortfeld T, van Herk M and Jiang S B 2002d When should systematic patient positioning
errors in radiotherapy be corrected? Phys. Med. Biol. 47 N297302
Bos L J, de Boer R W, Mijnheer B J, Lebesque J V and Damen E M F 2001 Forward versus
inverse planning of static IMRT of prostate cancer Radiother. Oncol. 61 (Suppl. 1)
S32
Bos L J, Schwarz M, Bar W, Alber M, Mijnheer B J, Lebesque J V and Damen E M F
2004 Comparison between manual and automatic segment generation in step-andshoot IMRT of prostate cancer Med. Phys. 31 12230
References
389
Bouchet L G, Meeks S L, Bova F J, Goodchild G, Buatti J M and Friedman W A 2000

Three-dimensional ultrasound optic guidance for high precision radiation therapy
Proc. 42nd Annual ASTRO Meeting. Int. J. Radiat. Oncol. Biol. Phys. 48 (Suppl.)
195
Boyer A 2001 Clinical implementation of IMRT Med. Phys. 28 1222
2003 Static MLC IMRT (Step and shoot) IMRTthe State of the Art: AAPM Medical
Physics Monograph Number 29 ed J R Palta and T R Mackie (Madison, WI: Medical
Physics Publishing) pp 285317
Boyer A, Song Y, Yong D and Xing L 2002a Sweeping window arc therapy (SWAT) Med.
Phys. 29 1265
Boyer A, Xing L, Luxton G and Ma C 2000 IMRT with dynamic multileaf collimation
Boyer A, Xing L, Yu C, Xia P and Verhey L 2002b Linear accelerator quality assurance
for IMRT Radiother. Oncol. 64 (Suppl. 1) S85
Braaksma M, Levendag P, Muller K, Van de Est H, Van Sornsen de Koste J and Nowak P
2002 The role of IMRT in advanced cancers of the head and neck. Some limitations
Braaksma M M J, Wijers O B, van Sornsen de Koste J R, van der Est H, Schmitz P I M,
Nowak P J C M and Levendag P C 2003 Optimisation of conformal radiation therapy
by intensity modulation: cancer of the larynx and salivary gland function Radiother.
Oncol. 66 291302
Bradley J, Thorstad W L, Mutic S, Miller T R, Dehdashti F, Siegel B A, Bosch W and
Bertrand R J 2004 Impact of FDG-PET on radiation therapy volume delineation in
non-small-cell lung cancer Int. J. Radiat. Oncol. Biol. Phys. 59 7886
Bragg C M, Conway J and Robinson M H 2001 Parotid gland tumoursA comparison
of intensity modulated radiotherapy and 3-D conformal radiotherapy Int. J. Radiat.
Oncol. Biol. Phys. 51 (Suppl. 1) 3423
Brahme A 1988 Optimisation of stationary and moving beam radiation therapy techniques
Radiother. Oncol. 12 12940
Brahme A, Roos J E and Lax I 1982 Solution of an integral equation encountered in rotation
therapy Phys. Med. Biol. 27 12219
Brock K K, Sharpe M B and Jaffray D A 2004 Biomechanical modelling of anatomical
deformation for image-guided registration of soft-tissue surrogates Proc. 14th Int.
Conf. on the Use of Computers in Radiation Therapy (Seoul, May) pp 46771
Broggi S, Fiorino C, Castellone P, Bertelli E, Cattaneo G M and Calandrino R 2002
Mechanical testing and dosimetric characterization of a 1D IMRT delivering system
for clinical applications Radiother. Oncol. 64 (Suppl. 1) S2045
Bruinvis A D and Damen E M F 2001 Quality assurance of a treatment planning system
for IMRT applications Radiother. Oncol. 61 (Suppl. 1) S45
Bucciolini M, Buonamici F B and Casati M 2004 Verification of IMRT fields by film
dosimetry Med. Phys. 31 1618
Budgell G 1998 Delivery of intensity modulated radiation therapy by dynamic multileaf
collimation: resolution requirements for specifying leaf trajectories. Med. Phys. 25
A202
Budgell G 2002 Verification in intensity-modulated radiotherapy RAD Magazine 29 (323)
234
390
References
Budgell G J, Martens C and Claus F 2001a Improved delivery efficiency for step and shoot
intensity modulated radiotherapy using a fast-tuning magnetron Phys. Med. Biol. 46
N25361
Budgell G J, Mott J H L, Logue J P and Hounsell A R 2001b Clinical implementation of
dynamic multileaf collimation for compensated bladder treatments Radiother. Oncol.
59 318
Budgell G J, Sykes J R and Wilkinson J M 1998 Rectangular edge synchronization for
intensity modulated radiation therapy with dynamic multileaf collimation Phys. Med.
Biol. 43 276984
Burnet N 2003 Conformal therapy + rotation and central axis block is an alternative to
IMRT for spine tumours Clinical Oncol. 15 (Suppl. 2) S23
Buyyounouski M K, Horwitz E M, Price R A, Hanlon A L, Uzzo R G and Pollack A 2004
Intensity-modulated radiotherapy with MRI simulation to reduce doses received by
erectile tissue during prostate cancer treatment Int. J. Radiat. Oncol. Biol. Phys. 58
7439
Caccia B, Del Giudice P, Mattia M, Benassi M and Marzi S 2001 An evaluation of the
overall sensitivity of the optimisation process in the IMRT technique Radiother.
Cadman P, Bassalow R, Sidhu N P S, Ibbott G and Nelson A 2002 Dosimetric
considerations for validation of a sequential IMRT process with a commercial
treatment planning system Phys. Med. Biol. 47 300110
Caldwell C B, Mah K, Skinner M and Danjouz C E 2003 Can PET provide the 3D extent
of tumor motion for individualized internal target volumes? A phantom study of
the limitations of CT and the promise of PET Int. J. Radiat. Oncol. Biol. Phys. 55
138193
Cardarelli G, Zheng Z, Hussain K, Cintron O, Shearer D, DiPetrillo T, Tsai J, Engler M,
Mohiuddin and Wazer D 2001 Evaluation of a new commercial QA phantom for
intensity modulated radiation therapy (IMRT) plan verification Med. Phys. 28 1209
Caudrelier J, Vial S, Gibon D, Bourel P, Vasseur C and Rousseau J. 2000 Evaluation of a
fuzzy logic method for volume determination in MR imaging Radiother. Oncol. 56
(Suppl. 1) S17
Caudrelier J-M, Vial S, Gibon D, Kulik C, Fournier C, Castelain B, Coche-Dequeant B
and Rousseau J 2003 MRI definition of target volumes using fuzzy logic method for
three-dimensional conformal radiation therapy Int. J. Radiat. Oncol. Biol. Phys. 55
22533
Cavey M L, Bayouth J E, Colman M, Endres E J and Sanguineti G 2004 Is dose escalation
to the prostate feasible while treating the pelvic nodes with IMRT? Proc. 14th Int.
Chandra A, Dong L, Huang E, Kuban D A, ONeill L J, Rosen I I and Pollack A 2001
Evaluation of ultrasound-based daily prostate localization during IMRT For prostate
cancer Int. J. Radiat. Oncol. Biol. Phys. 51 (Suppl. 1) 1678
2003 Experience of ultrasound-based daily prostate localization Int. J. Radiat. Oncol.
Biol. Phys. 56 43647
Chang J 2001 IMRT (intensity modulated radiotherapy) verification using Varian AS500
electronic portal imaging device (EPID) Med. Phys. 28 12823
Chang J, Mageras G K, Chui C S, Ling C C and Lutz W 2000a Relative profile and dose
verification of intensity-modulated radiation therapy Int. J. Radiat. Oncol. Biol. Phys.
47 23140
References
391
Chang S X, Cullip T J and Deschesne K M 2000b Intensity modulation delivery techniques:

step and shoot MLC auto-sequence versus the use of a modulator Med. Phys. 27
94859
Chang S, Cullip T, Deschesne K, Miller E and Rosenman J 2002 Clinical experience of
an alternative IMRT delivery technique intensity modulated via compensators Med.
Phys. 29 1305
2004a Compensators: An alternative IMRT delivery technique Proc. 14th Int. Conf.
on the Use of Computers in Radiation Therapy (Seoul, May) pp 20711
Chang S, Cullip T, Deschesne K and Rosenman J 2003 Clinical experience of a
compensator-based intensity modulated treatment technique Proc. World Congress
in Medical Physics (Sydney) CDROM and website only
Chang S, Potter L and Cullip T 2001 Investigation of a camera-type EPID as an IMRT QA
tool Med. Phys. 28 1293
2003 High intensity modulation resolution: the advantage of a 4-bank micro-multileaf
collimator system Proc. World Congress in Medical Physics (Sydney) CDROM and
website only
Chang S, Schulman D, Cullip T, Keall P, Mageras G and Joshi S 2004b The effect of
organ motion on cumulative dosimetry: a comparison of different IMRT delivery
techniques Proc. 14th Int. Conf. on the Use of Computers in Radiation Therapy
(Seoul, May) pp 61517
Chao C and Blanco A L 2003 IMRT for head and neck cancer IMRTthe State of the
Art: AAPM Medical Physics Monograph Number 29 ed J R Palta and T R Mackie
(Madison, WI: Medical Physics Publishing) pp 63144
Chao C, Deasy J O, Markman J, Haynie J, Perez C A, Purdy J A and Low D A 2000
Functional outcome of parotid gland sparing in patients with head and neck (H and N)
cancers receiving intensity-modulated (IMRT) or 3-D radiation therapy Proc. 42nd
Annual ASTRO Meeting. Int. J. Radiat. Oncol. Biol. Phys. 48 (Suppl.) 174
Chao K S C, Majhail N, Huang C-J, Simpson J R, Perez C A, Haughey B and
Spector G 2001 Intensity-modulated radiation therapy reduces late salivary toxicity
without compromising tumor control in patients with oropharyngeal carcinoma: a
comparison with conventional techniques Radiother. Oncol. 61 27580
Chapman J D 2003 How can PET contribute to our current RT treatment planning?
Chapman J D. Movsas B and Nahum A E 2003 Fifty years later: how many current
treatment failures are due to hypoxia? Radiother. Oncol. 68 (Suppl. 1) S32
Chauvet I, Papatheodorou S, Gaboriaud G, Ponvert D and Rosenwald J C 2001 Preliminary
study of optimization parameters for the HELIOS inverse planning software Phys.
Medica. 17 115
Chen G T, Jiang S B, Kung J, Doppke K P and Willett C G 2001b Abdominal organ motion
and deformation: Implications for IMRT Int. J. Radiat. Oncol. Biol. Phys. 51 (Suppl.
1) 210
Chen H, Yang J, Stapleton P and Murphy S 2002a Calibration of MLC leaf positions with
an electronic portal imaging device for IMRT delivery Med. Phys. 29 1262
Chen J Z and Wong E 2004 MLC leaf optimisation for intensity modulated arc therapy
Proc. 14th Int. Conf. on the Use of Computers in Radiation Therapy (Seoul, May)
pp 1324
Chen L, Price R, Li J, Wang L, Qin L, Ma C and Pollack A 2003 MRI-based treatment
optimisation for prostate IMRT Med. Phys. 30 1335
392
References
Chen Q-S, Weinhouse M S, Deibel F C, Ciezki J P and Macklis R M 2001a Fluoroscopic

study of tumour motion due to breathing: facilitating precise radiation therapy for
lung cancer patients Med. Phys. 28 18506
Chen Y, Boyer A L and Ma C-M 2000a Calculation of x-ray transmission through a
multileaf collimator Med. Phys. 27 171726
Chen Y and Galvin J 2001 Optimization of conformal radiotherapy by beam weight
renormalization Med. Phys. 28 1262
Chen Y, Hou Q and Galvin J M 2004 A graph-searching method for MLC leaf sequencing
under constraints Med. Phys. 31 150411
Chen Y, Michalski D, Houser C and Galvin J M 2002b A deterministic iterative leastsquares algorithm for beam weight optimization in conformal radiotherapy Phys.
Med. Biol. 47 164758
Chen Y, Michalski D, Xiao Y and Galvin J M 2001c Automatic aperture selection and
IMRT plan optimization by beam weight renormalization Int. J. Radiat. Oncol. Biol.
Phys. 51 (Suppl. 1) 745
Chen Y, Xing L, Luxton G, Li J L and Boyer A L 2000b Automated quality asurance
process for MLC-based IMRT using CT virtual simulation Proc. 42nd Annual ASTRO
Meeting. Int. J. Radiat. Oncol. Biol. Phys. 48 (Suppl.) 194
Cheng C-W and Das I J 2002 Comparison of beam characteristics in intensity modulated
radiation therapy (IMRT) and those under normal treatment condition. Med. Phys. 29
22630
Cheng C W, Das I J and Huq M S 2003 Lateral loss and dose discrepancies of multileaf
collimator segments in intensity modulated radiation therapy Med. Phys. 30 295968
Cheng C, Das I, Xiao Y and Galvin J 2002 Beam characteristics in intensity modulated
radiation therapy (IMRT) of modern accelerators Med. Phys. 29 1266
Cheng C W, Wong J R, Ndlovu A M, Das I J, Schiff P and Uematzu M 2000 Dosimetric
study of virtual mini-multileaf and its clinical application Proc. 42nd Annual ASTRO
Cho B C J, Hurkmans C W Damen E M F and Mijnheer B J 2001a A planning
study comparing intensity modulated versus non-intensity modulated radiotherapy
in treating breast cancer Radiother. Oncol. 61 (Suppl. 1) S44
Cho B C J, Hurkmans C W, Damen E M F, Zijp L J and Mijnheer B J 2002a Intensity
modulated versus non-intensity modulated radiotherapy in the treatment of the
left breast and upper internal mammary lymph node chain: a comparative study
Cho B, Schwarz M, Mijnheer B and Bartelink H 2002b A simple intensity
modulated radiotherapy class solution using predefined segments to reduce cardiac
complications in the treatment of left-sided breast cancer Med. Phys. 29 1284
2004 Simplified intensity modulated radiotherapy class solution using pre-defined
segments to reduce cardiac complications in left-sided breast cancer Radiother.
Oncol. 70 23141
Cho J 2002 Intensity modulated radiotherapy in breast cancer Radiother. Oncol. 64 (Suppl.
1) S89
Cho P S and Marks II R J 2000 Hardware-sensitive optimisation for intensity modulated
radiotherapy. Phys. Med. Biol. 45 42940
Cho P, Meyer J, Yee D, Phillips M and Parsai H 2003 Numerical instability and
conditioning in IMRT optimization Med. Phys. 30 1489
References
393
Cho P S, Parsaei H and Phillips M H 2001b Influence of random field perturbations on

dynamic intensity modulated beams Med. Phys. 28 12034
Cho P S and Phillips M H 2001 Reduction of computational dimensionality in inverse
radiotherapy planning using sparse matrix operations Phys. Med. Biol. 46 N11725
Choi B and Deasy J O 2002 The generalized equivalent uniform dose function as a basis
for intensity-modulated treatment planning Phys. Med. Biol. 47 357989
Christian J A, McNair H A, Symonds-Tayler J R N, Knowles C, Bedford J L and Brada
M 2003a Reduction in lung tumour motion using active breathing control Radiother.
Christian J A, Partridge M, Cook G, McNair H A, Cronin B, Courbon F, Bedford J
L and Brada M 2003b Optimisation of lung radiotherapy planning using accurate
co-registration of planning CT and SPECT perfusion images Radiother. Oncol. 68
(Suppl. 1) S54
Chu S and Suh C 2000 Dose planning of intensity modulated radiotherapy for
nasopharyngeal cancer using compensating filters Proc. World Congress of Medical
Physics and the AAPM Annual Congress, August 2000 paper WE-E313-09
Chuang C F, Verhey L J and Xia P 2002 Investigation of the use of MOSFET for clinical
IMRT dosimetric verification Med. Phys. 29 1109115
Chuang C, Xia P, Nguyen-Tan F, Fu K and Verhey L 2000 Investigation of the uncertainties
in patient positioning and patient motion in IMRT treatment Proc. World Congress of
Medical Physics and the AAPM Annual Congress, August 2000 paper WE-FXH-45
Chuang K-S, Chen T-J, Kuo S-C, Jan M-L, Hwang I-M, Chen S, Ling Y-C and Wu J 2003
Determination of beam intensity in a single step for IMRT inverse planning Phys.
Med. Biol. 48 293306
Chui C 2000 Clinical implementation of IMRT Proc. World Congress of Medical Physics
and the AAPM Annual Congress, August 2000 paper TU-ABR-02
Chui C S, Chan M F and Ling C C 2000b Delivery of intensity-modulated radiation therapy
with a multileaf collimator: Comparison of step-and-shoot and dynamic leaf motion
methods Proc. World Congress of Medical Physics and the AAPM Annual Congress,
August 2000 paper MO-EBR-06
Chui C-S, Chan M F, Yorke E, Spirou S and Ling C C 2001a Delivery of intensitymodulated radiation therapy with a conventional multileaf collimator: Comparison
of dynamic and segmental methods Med. Phys. 28 24419
Chui C-S, Hong L and Hunt M 2002 A simplified intensity modulated radiation therapy
technique for the breast Med. Phys. 29 5229
Chui C, Hong L, Yang J and Spirou S 2001b Splitting of large intensity-modulated fields
Med. Phys. 28 1252
Chui C, LoSasso T and Palm A 2003a Computational and empirical verification of
intensity modulated radiation therapy delivered with a multileaf collimator Proc.
World Congress in Medical Physics (Sydney) CDROM and website only
Chui C-S, Spirou S and LoSasso T 1996 Testing of dynamic multileaf collimation Med.
Phys. 23 63541
Chui C-S, Yorke E and Hong 2003b The effects of intra-fraction organ motion on the
delivery of intensity-modulated field with a multileaf collimator Med. Phys. 30 1736
46
Chytka T, Novotny J Jr, Vymazal J, Liscak R, Marek J and Vladyka V 2000 Glioma
malignancy prediction model based on patients clinical data and CT and MRI
imaging Radiother. Oncol. 56 (Suppl. 1) S37
394
References
Ciernik I F, Dizendorf E, Reiner B, Burger C, Khan S, Lutolf U M, Steinert H,

VonSchuthess G K and Baumert B G 2002 Treatment planning of 3D-conformal
radiation therapy based on the integrated computer-assisted Positron Emission
Tomography (PET/CT)imaging Radiother. Oncol. 64 (Suppl. 1) S92
Clark C, Bidmead M, Mubata C, Harrington K, Evans P R and Nutting C 2003a Does
IMRT provide an improved treatment for carcinoma of the larynx and cervical nodes
Clinical Oncol. 15 (Suppl. 2) S5
Clark C H, Bidmead M A, Mubata C D, Harrington K J and Nutting C M 2004 Intensitymodulated radiotherapy improves target coverage, spinal cord sparing and allows
dose escalation in patients with locally advanced cancer of the larynx Radiother.
Oncol. 70 18998
Clark C, Bidmead M, Mubata C, Harrington K, Rhys Evans P and Nutting C 2002c
Planning and quality assurance issues for clinical implementation of intensity
modulated radiotherapy (IMRT) for treatment of carcinoma of the larynx and cervical
nodes Radiother. Oncol. 64 (Suppl. 1) S2456
Clark C, Hansen V N, Miles E, McNair H, Bidmead M, Webb S and Dearnaley D 2003c
Optimization of forward planned IMRT for hypofractionated treatment of the prostate
Clark C H, Miles E A, Bidmead M A, Mubata C D, Harrington K J and Nutting C M 2003d
(in regard to Lee et al 2002 Int. J. Radiat. Oncol. Biol. Phys. 53 6307) Int. J. Radiat.
Clark C, Miles E, Mubata C, Meehan C, Alonso S and Bidmead M 2003b Optimization of
a patient specific quality assurance procedure for dynamic IMRT Clinical Oncol. 15
(Suppl. 2) S7
Clark C H. Mubata C D, Bidmead A M, Humphreys M, McNair H, Harrington K J
and Nutting C M 2002b Clinical implementation of dynamic intensity modulated
radiotherapy: Planning and quality assurance issues for treatment of patients
with carcinoma of the thyroid and cervical nodes Proc. IPEM Biennial Meeting
(Southampton, 1617 July) pp 910
Clark C H, Mubata C D, Meehan C A, Bidmead A M, Staffurth J, Humphreys M E
and Dearnaley D P 2002a IMRT clinical implementation: prostate and pelvic node
irradiation using HELIOS and a 120-leaf MLC J. Appl. Clinical Med. Phys. 3 27384
Claus F, Boterberg T, Ost P and de Neve W 2002 Short term toxicity profile for 32 sinonasal
cancer patients treated with IMRT. Can we avoid dry eye syndrome? Radiother.
Oncol. 64 2058
Claus F, de Gersem W, Vanhoutte I, Duthoy W, Remouchamps V, de Wagter C and de
Neve W 2001 Evaluation of a leaf position optimization tool for intensity modulated
radiation therapy of head and neck cancer Radiother. Oncol. 61 2816
Claus F, de Wagter C, Remouchamps V, de Gersem W and de Neve W 2000 Time efficiency
and monitor unit efficiency of anatomy-based IMRT Radiother. Oncol. 56 (Suppl. 1)
S195
Convery D J 2001 Implementation of inverse-planned IMRT with dynamic MLC delivery
at the Royal Marsden NHS Trust, Sutton Compendium of the 3rd IMRT Winter
School, Heidelberg (London, Jan 26/27) pp D2.1D2.7
Convery D J, Cosgrove V P and Webb S 2000 Improving dosimetric accuracy of a dynamic
MLC technique Proc. 13th Int. Conf. on the Use of Computers in Radiation Therapy
(Heidelberg, May) ed W Schlegel and T Bortfeld (Heidelberg: Springer) pp 2779
References
395
Convery D J and Webb S 1997 Calculation of the distribution of head-scattered radiation

in dynamically collimated MLC fields Proc. 12th Int. Conf. on the Use of Computers
in Radiation Therapy (Salt Lake City, UT, May) ed D D Leavitt and G Starkshall
1998 Generation of discrete beam-intensity modulation by dynamic multileaf
collimation under minimum leaf separatation constraints Phys. Med. Biol. 43 2521
38
Coolens C, Webb S, Evans P M and Seco J 2003 Combinational use of conformal and
intensity modulated beams in radiotherapy planning Phys. Med. Biol. 48 1795807
Cooper P, MacKay R I and Williams P C 2001a Dosimetric evaluation of a resolutionenhancing tertiary collimator for radiotherapy Radiother. Oncol. 61 (Suppl. 1) S57
2001b A resolution-enhancing tertiary collimator for radiotherapy: artefacts
generated using the protytype and a strategy to avoid them Radiother. Oncol. 61
(Suppl. 1) S81
2001c Optimizing the indexing of a resolution-enhancing tertiary collimator for
radiotherapy Phys. Med. Biol. 46 N2638
2002 Indexing artefacts using a tertiary collimator and a method to avoid them Phys.
Med. Biol. 47 N191201
Corletto D 2001 Inverse and forward optimisation by 1D and 2D IMRT for prostate cancer
Cosgrove V 2001 Verification and QA of IMRT by DMLC technique Compendium of the
3rd IMRT Winter School, Heidelberg (London, Jan 26/27) pp V1.1V1.8
Cosgrove V P, Carson K J, Hounsell A R, Grattan M W D, Stewart D P, Eakin R L, Fleming
V A L and Jarritt P H 2003 Direct use of an integrated PET/CT imaging system for
radiotherapy treatment planning in non-small-cell lung cancer Radiother. Oncol. 68
(Suppl. 1) S29
Cosgrove V P, Convery D J, McNair H A, Vaarkamp J and, Webb S 2001 Quality assurance
in a phase 1 prostate and pelvic node IMRT trial Phys. Medica. 17 1601
Cosgrove V P, Convery D J, Murphy P S, Nutting C M and Webb S 2000 Dynamic
MLC delivered IMRT: verification using polyacrylamide gel dosimetry Proc. 13th
Int. Conf. on the Use of Computers in Radiation Therapy (Heidelberg, May) ed
W Schlegel and T Bortfeld (Heidelberg: Springer) pp 31113
Cotrutz C 2002 Inverse treatment planning with interactively variable voxel-dependent
importance factors Med. Phys. 29 1336
Cotrutz C, Kappas C and Webb S 2000 Intensity modulated arc therapy (IMAT) with
centrally blocked rotational fields Phys. Med. Biol. 45 2185206
Cotrutz C and Xing L 2002 Using voxel-dependent importance factors for interactive DVH
based optimization Phys. Med. Biol. 47 165969
2003a IMRT dose shaping with regionally variable penalty scheme Med. Phys. 30
54451
2003b Segment-based IMRT inverse planning using a genetic algorithm Med. Phys.
30 1372
2003c Segment-based dose optimisation using a genetic algorithm Phys. Med. Biol.
48 298798
Cozzi L and Fogliata A 2001 Dosimetric validation of a step and shoot intensity modulation
technique using ionisation chambers, films and a commercial electronic portalimaging device Radiother. Oncol. 61 (Suppl. 1) S107
396
References
Cozzi L, Fogliata A and Bolsi A 2001a A treatment planning comparison of 3D conformal

therapy, intensity modulated photon therapy and proton therapy for treatment of
advanced head and neck tumours Radiother. Oncol. 61 (Suppl. 1) S32
2001b Comparative analysis of IMRT step and shoot and sliding window techniques
on two commercial treatment planning modules Radiother. Oncol. 61 (Suppl. 1) S65
2001c Intensity modulation for intact breast cancer treatment. A first approach
Cozzi L, Fogliata A , Bolsi A, Nicolini G and Bernier J 2004 Three-dimensional conformal
vs intensity-modulated radiotherapy in head-and-neck cancer patients: comparative
analysis of dosimetric and technical parameters Int. J. Radiat. Oncol. Biol. Phys. 58
61724
Cozzi L, Fogliata A, Lomax A and Bolsi A 2001d A treatment planning comparison of
3D conformal therapy, intensity modulated photon therapy and proton therapy for
treatment of advanced head and neck tumours Radiother. Oncol. 61 28797
Crooks S M, McAven L F, Robinson D R and Xing L 2000 Treatment time reduction
through multileaf-collimator leaf-sequencing in static IMRT Proc. 42nd Annual
2002 Minimizing delivery time and monitor units in static IMRT by leaf-sequencing
Phys. Med. Biol. 47 310516
Crooks S M, Wu X, Takita C, Watzich M and Xing L 2003 Aperture modulated arc therapy
Crooks S M and Xing L 2001 Linear algebraic methods applied to intensity modulated
radiation therapy Phys. Med. Biol. 46 2587606
Crosbie J C, Poynter A J, James H V, Scrase C D, Morgan J S, Bulusu V R and LeVay J
2002 Quality assurance program for effective delivery of dynamic IMRT Proc. IPEM
Biennial Meeting (Southampton, 1617 July) p 36
Cumberlin R L, Deye J and Coleman C N 2002 In response to Drs Schulz and Kagan Int.
Curran B 2003 IMRT delivery using serial tomotherapy IMRTthe State of the Art: AAPM
Dai J-R and Hu Y-M 1999 Intensity-modulation radiotherapy using independent
collimators: an algorithm study Med. Phys. 26 256270
Dai J and Zhu Y 2001a Minimizing the number of segments in a delivery sequence for
intensity-modulated radiation therapy with a multileaf collimator Med. Phys. 28
12512
2001b Minimizing the number of segments in a delivery sequence for intensitymodulated radiation therapy with a multileaf collimator Med. Phys. 28 211320
Daisne J F, Duprez T, Weyant B, Sibomana M, Reychler H, Hamoir M, Lonneux M,
Cosnard G and Gregoire V 2002 Impact of computed tomography (CT), magnetic
resonance (MR) and positron emission tomography with fluorodeoxyglucose (FDGPET) image coregistration on GTV delineation in head and neck squamous cell
carcinoma (HNSCC) Radiother. Oncol. 64 (Suppl. 1) S256
Danciu C and Proimos B 2003 Gravity oriented absorbers for conformal therapy of small
and irregular targets with protection of their vital neighbours Proc. World Congress
Das S, Bell M, Zhou S, Miften M, Munley M, Whiddon C, Craciunescu O, Baydush A,
Wong T, Rosenman J, Dewhirst M and Marks L 2003 Feasibility of functional image-
References
397
guided radiotherapy evaluated using a novel heuristic fluence modulation algorithm

Med. Phys. 30 1335
Das S, Cullip T, Tracton G, Chang S, Marks L, Anscher M and Rosenman J 2003
Beam orientation selection for intensity-modulated radiation therapy based on target
equivalent uniform dose maximization Int. J. Radiat. Oncol. Biol. Phys. 55 21524
Das S, Halvorsen P, Cullip T, Chang S, Tracton G, Marks L and Rosenman J 2001a
Equivalent uniform dose based automated beam selection for intensity modulated
treatment planning Med. Phys. 28 1307
Das S K, Halvorsen P, Cullip T, Tracton G, Chang S, Marks L, Anscher M and Rosenman
J 2001b Beam orientation selection for intensity modulated radiation therapy based
on target equivalent uniform dose maximization Int. J. Radiat. Oncol. Biol. Phys. 51
74
Das S, Miften M, Zhou S, Bell M, Munley M, Whiddon C, Craciunescu O, Baydush
A, Wong T, Rosenman J, Dewhirst M and Marks L 2004 Feasibility of optimising
the dose distribution in lung tumours using fluorine-18-fluorodeoxyglucose positron
emmission tomography and single photon emission computed tomography guided
dose prescriptions Med. Phys. 31 145261
Dasu A, Toma-Dasu I and Fowler J F 2003 Should single or distributed parameters be used
to explain the steepness of tumour control probability curves? Phys. Med. Biol. 48
38797
Dawson L A, Brock K, Litzenberg D, Kazanjian S, McGinn C G, Lawrence T S, Ten Haken
R K and Balter J 2000 The reproducibility of organ position using active breathing
control (ABC) during liver radiotherapy Proc. 42nd Annual ASTRO Meeting. Int. J.
Radiat. Oncol. Biol. Phys. 4 (Suppl.) 1675
Dearnaley D P 2000 Optimisation of high dose treatment for prostate cancer from
conformal to IMRT Radiother. Oncol. 56 (Suppl. 1) S123
2004 (in regard to 2003 The return of the snake oil salesman Int. J. Radiat. Oncol.
Biol. Phys. 55 5612 and 56 15078) Int. J. Radiat. Oncol. Biol. Phys. 58 1644
Dearnaley D P, Khoo V S, Norman A R, Meyer L, Nahum A, Tait D, Yarnold J and
Horwich A 1999 Comparison of radiation side effects of conformal and conventional
radiotherapy in prostate cancer: a randomised trial Lancet 353 26772
Deasy J and El Naqa I 2003 Adaptive gridding for IMRT dose calculations Med. Phys. 3
1487
Deasy J, Lee E, Kawrakow I and Zakarian C 2002 A prototype Monte Carlo-based IMRT
treatment planning research system Med. Phys. 29 1254
Deasy J and Wickerhauser M 2001 IMRT optimization based on stored pencil beam dose
distributions: Compression, denoising and dose calculations using wavelets Med.
Phys. 28 1261
Deasy J and Zakarian C 2003 Dose-distance constraints: A new method for controlling
intensity modulated treatment planning dose distributions Med. Phys. 30 1334
De Brabandere M, van Esche A, Kutcher G J and Huyskens D 2002 Quality assurance in
intensity modulated radiotherapy by identifying standards and patterns in treatment
preparation: a feasibility study on prostate patients Radiother. Oncol. 62 28391
Debus J, Pirzkall A, Thilmann C, Grosser K H, Rhein B, Haering P, Hoess A and
Wannenmacher M 2000 IMRT in the treatment of skull base tumours Radiother.
398
References
De Deene Y, de Wagter C and Baldock C 2003 Comment on A systematic review of the

precision and accuracy of dose measurements in photon radiotherapy using polymer
and Fricke MRI gel dosimetry Phys. Med. Biol. 48 L1518
De Deene Y, de Wagter C and de Neve W 2000 Monomer/polymer gel dosimetry
De Gersem W, Claus F, de Wagter C and de Neve W 2001a An anatomy-based beam
segmentation tool for intensity-modulated radiation therapy and its application to
head-and-neck cancer Int. J. Radiat. Oncol. Biol. Phys. 51 84959
De Gersem W, Claus F, de Wagter C, Van Duse B and de Neve W 2001b Leaf position
optimization for step-and-shoot IMRT Int. J. Radiat. Oncol. Biol. Phys. 51 137188
De Gersem W, Vakaet L, Claus F, Remouchamps V, Van Duyse B and de Neve W 2000
MLC leaf position optimization using a biophysical objective function Radiother.
Dehdashti F, Grigsby P W, Mintun M A, Lewis J S, Siegel B A and Welch M J 2003
Assessing tumor hypoxia in cervical cancer by positron emission tomography with
60 Cu-ATSM: relationship to therapeutic response-a preliminary report Int. J. Radiat.
Dehnad H 2001 Clinical use of IMRT for prostate treatment with marker detection
Abstracts of the Elekta IMRT User Meeting 2001 in NKI Amsterdam pp 1820
Dejean C, Grandjean P, Lefkopoulos D and Foulquier J 2000b Application of regularisation
methods to the IMRT inverse problem Radiother. Oncol. 56 (Suppl. 1) S211
Dejean C, Lefkopoulos D, Grandjean P, Berre F and Touboul E 2000c Comparison of
two inversion techniques: the analytic singular value decomposition and the iterative
gradient with constraints Radiother. Oncol. 56 (Suppl. 1) S212
Dejean C, Lefkopoulos D, Platoni K, Keraudy K and Julia F 2000a ROC analysis: a useful
evaluation tool to define the most conformal prescription isodose for IMRT plans
Dejean C, Lemosquet A, Grandjean P, Lefkopoulos D and Touboul E 2001a Influence
of IMRT inverse problem regularisation on biological indices Radiother. Oncol. 61
(Suppl. 1) S84
Dejean C, Lemosquet A, Lefkopoulos D and Touboul E 2001b Comparison of biological
assessment functions in intensity modulated radiotherapy: two-dimensional study
2001c Comparison of biological assessment functions in intensity modulated
radiotherapy: A two-dimensional study Phys. Medica. 17 115
De Langen M, Essers M, Dirkx M L P, Heijimen B J M 2003 Bringing IMRT into clinical
practice: The complete trajectory Phys. Medica. 19 59
Della Biancia C, Hunt M and Amols H 2002 A comparison of the integral dose from 3D
conformal and IMRT techniques in the treatment of prostate cancer Med. Phys. 29
1216
Della Biancia C, Yorke E, Mageras G, Giraud P, Rosenzweig K, Amols H and Ling C 2003
Comparison of two breathing levels for gated intensity modulated radiation therapy
(IMRT) of lung cancer Med. Phys. 30 23401
Deloar H M, Kunieda E, Kawase T, Saitoh H, Tohyama N, Ozaki M, Kimura H, Fujisaki T
and Kubo A 2004 Tomotherapy with kilo-voltage X-ray energy: Preliminary trials
with Monte Carlo simulation Proc. 14th Int. Conf. on the Use of Computers in
Radiation Therapy (Seoul, May) pp 1947
References
399
De Meerleer G, de Neve W, Verbaeys A, Goethals C, Villeirs G, Oosterlinck W and Vakaet

L 2002a Intensity modulated radiotherapy (IMRT) for prostate cancer: acute toxicity
of 102 patients treated Radiother. Oncol. 64 (Suppl. 1) S56
De Meerleer G O, Vakaet L A M L, de Gersem W R T, de Wagter C, de Naeyer B and de
Neve W 2000a Radiotherapy of prostate cancer with or without intensity modulated
beams: a planning comparison Int. J. Radiat. Oncol. Biol. Phys. 47 63948
De Meerleer G, Vakaet L and de Neve W 2000b Three versus five beams in the IMRT
treatment for prostate cancer: a planning comparison Radiother. Oncol. 56 (Suppl. 1)
S107
De Meerleer G, Villeirs G, Bral S, de Gersem W and de Neve W 2002b Simultaneous
boost to the intraprostatic lesions: IMRT planning for step and shoot delivery by
direct segment aperture optimization Radiother. Oncol. 64 (Suppl. 1) S286
Dempsey J F, Ahuja R K, Kamath S, Kumar A, Li J G, Palta J R, Romeijn H E, Ranka S
and Sahni S 2003 The leaf sequencer: an underestimated problem? Radiother. Oncol.
68 (Suppl. 1) S3
Deng J, Guerrero T, Ding M S, Jolly J, Pawlicki T and Ma C M 2001c Incorporate organ
motion into MLC leaf sequencing for intensity modulated radiation therapy Int. J.
Radiat. Oncol. Biol. Phys. 51 (Suppl. 1) 923
Deng J, Guerrero T, Hai J, Luxton G and Ma C-M 2001b Source modelling and beam
commissioning for a Cyberknife system Med. Phys. 28 1303
Deng J, Lee M, Ding M and Ma C 2002 Effect of breathing motion on IMRT treatment of
breast cancer: a Monte Carlo study Med. Phys. 29 1263
Deng J, Pawlicki T, Chen Y, Li J, Jiang S and Ma C 2000 The MLC tongue-and-groove
effect on IMRT dose distributions Proc. World Congress of Medical Physics and the
AAPM Annual Congress, August 2000 paper WE-E313-06
2001a The MLC tongue-and-groove effect on IMRT dose distributions Phys. Med.
Biol. 46 103960
Depuydt T, Van Esche A and Huyskens D P 2002 A quantitative evaluation of IMRT dose
distributions: refinement and clinical assessment of the gamma evaluation Radiother.
Oncol. 62 30919
Deshpande N and Poynter A 2000 Effect of differential field margins in conformal therapy
of the prostate Radiother. Oncol. 56 (Suppl. 1) S107
De Wagter C, de Deene Y, Vergote K, Martens C, de Gersem W, Van Duyse B and de Neve
W 2001 The entire IMRT delivery process Radiother. Oncol. 61 (Suppl. 1) S16
Didinger B H, Nikoghosyan A, Schulz-Ertner D, Zierhut D, Schlegel W, Wannenmacher
M and Debus J 2001 Benefit of adaptive inverse planning during intensity modulated
radiotherapy of prostate cancera comparative DVH-analysis Int. J. Radiat. Oncol.
Dirkx M L P, de Langen M and Heijmen B J M 2000 Comparison of two algorithms for leaf
trajectory calculation for dynamic multileaf collimation (DMLC) Radiother. Oncol.
56 (Suppl. 1) S96
Dirkx M L P and Heijmen B J M 2000a Beam intensity modulation for penumbra
enhancement and field length reduction in lung cancer treatments: a dosimetric study
2000b Testing of the stability of intensity modulated beams generated with dynamic
multileaf collimation applied to the M50 Racetrack Microtron Med. Phys. 27 27017
Djajaputra D, Wu Q, Wu Y and Mohan R 2003 Algorithm and performance of a clinical
IMRT beam-angle optimisation system Phys. Med. Biol. 48 3191212
400
References
Dobbs H J 2000 ICRU-62: A clinical perspective Radiother. Oncol. 56 (Suppl. 1) S51

Dogan N, King S and Emami B 2003 Sequential versus simultaneous integrated boost
IMRT for treatment of head and neck cancers Radiother. Oncol. 68 (Suppl. 1) S74
Dogan N, Leybovich L B, King S and Sethi A 2001b Comparison of static step-and-shoot
IMRT plans developed by two different commercial treatment planning systems:
NOMOS-CORVUS vs CMS-FOCUS Radiother. Oncol. 61 (Suppl. 1) S82
Dogan N, Leybovich L B, King S, Sethi A and Emami B 2000c Comparison of treatment
plans developed for concave-shaped head and neck tumors: IMRT versus 3-D
conformal versus 2-D Proc. 42nd Annual ASTRO Meeting. Int. J. Radiat. Oncol. Biol.
Phys. 48 (Suppl. 1) 176
Dogan N, Leybovich L B and Sethi A 2001a Investigation of dose distributions in adjacent
static step-and-shoot IMRT and electron fields Radiother. Oncol. 61 (Suppl. 1) S66
2002b Investigation of split IMRT fields for step-and-shoot-IMRT delivery Radiother.
2003b An automatic feathering technique for split IMRT fields Med. Phys. 30 13356
Dogan N, Leybovich L, Sethi A and Emami B 2000b Comparison of IMRT plans for
treatment of head and neck cancers using modified and unmodified tomographic and
MLC-based static step-and-shoot techniques Radiother. Oncol. 56 (Suppl. 1) S106
2003a Automatic feathering of split fields for step-and-shoot intensity modulated
radiation therapy Phys. Med. Biol. 48 113340
2002a Improvement of dose distributions in abutment regions of intensity modulated
radiation therapy and electron fields Med. Phys. 29 3844
Dogan N, Leybovich L B, Sethi A, Krasin M and Emami B 2000a A modified method of
planning and delivery for dynamic multileaf collimator intensity-modulated radiation
therapy Int. J. Radiat. Oncol. Biol. Phys. 47 2415
Donaldson S S, Boyer A L and Hendee W R 2000 New methods for precision radiation
therapy exceed biological and clinical knowledge and institutional resources needed
for implementation Med. Phys. 27 24779
Dong D, Grant W, McGary J, Teh B and Butler E B 2000b Effect of beam energy in prostate
serial tomotherapy Proc. 42nd Annual ASTRO Meeting. Int. J. Rad. Oncol. Biol. Phys
48 (Suppl.) 347
Dong L and Court L 2003 Intercomparison of three in-room imaging modalities for patient
alignment: EPID, BAT, and CT-on-rails Med. Phys. 30 1473
Dong L, Court L, Wang H, ODaniel J and Mohan R 2004 Image-guided radiotherapy
with in-room CT-on-rails Proc. 14th Int. Conf. on the Use of Computers in Radiation
Dong L, Liu H, Wang X, Zhang X, Tu S, Mohan R and Wu Q 2003 The effect of photon
beam energy on the quality of IMRT plans Med. Phys. 30 1485
Dong L, McGary J, Bellezza D, Berner B and Grant W 2000a Whole-body dose from
Peacock-based IMRT treatment Proc. 42nd Annual ASTRO Meeting. Int. J. Rad.
Oncol. Biol. Phys 48 (Suppl.) 342
Donovan E M, Bleackley N J, Evans P M, Reise S F and Yarnold J R 2002b Dose-position
and dose-volume histogram analysis of standard wedged and intensity modulated
treatments in breast radiotherapy Brit. J. Radiol. 75 96773
Donovan E M, Johnson U, Shentall G, Evans P M, Neal A J and Yarnold J R 2000
Evaluation of compensators in breast radiotherapya planning study using multiple
static fields Int. J. Radiat. Oncol. Biol. Phys. 46 6719
References
401
Donovan E M, McNair H A, Eagle S, Shoulders B, Clements N, Evans P M, SymondsTayler J R N and Yarnold J R 2003 Minimizing cardiac irradiation in breast
radiotherapy with an automatic breathing control device: Initial findings Clinical
Donovan E M, McNair H, Evans P M, Symonds-Tayler R and Yarnold J R 2002a
Initial experience with an automatic breathing control device in breast radiotherapy
Drzymala R, Low D and Klein E 2001 Geometric aspects of intensity modulated radiation
therapy planning Med. Phys. 28 1205
Duan J, Shen S, Fiveash J, Brezovich I, Popple R and Pareek P 2002 Impact of dose
fractionation on dosimetric errors induced by respiratory motions during IMRT
delivery: a 3-D volumetric dose measurement approach Med. Phys. 29 1346
2003b Dosimetric effect of respiration-gated beam on IMRT delivery Med. Phys. 30
224152
Duan J, Shen S, Popple R, Ye S, Pareek P and Brezovich I 2003a Effect of the number of
treatment fields on respiration-induced dose errors in single and multi-fraction IMRT
delivery Med. Phys. 30 1340
Dyke L, Emery R, Lam T and Berson A 2001 Combining respiratory gating and forward
IMRT to treat patients with breast cancer of the left breast Med. Phys. 28 1987
Earl M A, Naqvi S, Shepard D M and Yu C X 2003c A comparison between direct aperture
optimization and two commercial IMRT systems Med. Phys. 30 1335
Earl M, Shepard D, Li X A, Naqvi S and Xu C X 2002 Simplifying IMRT with direct
aperture optimization Med. Phys. 29 1305
Earl M A, Shepard D M, Naqvi S, Li X A and Yu C X 2003a Inverse planning for intensitymodulated arc therapy using direct aperture optimization Phys. Med. Biol. 48 1075
89
Earl M A, Shepard D M, Naqvi S and Yu C X 2003b Jaws-only IMRT using direct aperture
optimization Med. Phys. 30 1487
Eberle K, Engler J, Hartmann G, Hofmann R and Horandel J R 2003 First tests of a liquid
ionisation chamber to monitor intensity modulated radiation beams Phys. Med. Biol.
48 355564
Eisbruch A 2002 Intensity-modulated radiotherapy of head-and-neck cancer: encouraging
early results Int. J. Radiat. Oncol. Biol. Phys. 53 13
El Naqa I, Low D A, Nystrom M, Parikh P, Lu W, Deasy J O, Amini A, Hubenschmidt J and
Wahab S 2004 An optical flow approach for automated breathing motion tracking in
4D computed tomography Proc. 14th Int. Conf. on the Use of Computers in Radiation
Engstrom P, Hansen H S, Haraldsson P, Landberg T, Engelholm S A and Nystrom H 2002
In vivo dose verification of IMRT treated head and neck patients Radiother. Oncol.
64 (Suppl. 1) S237
Erdi Y E, Yorke E D, Erdi A K, Braban L, Hu Y, Macapinlac H, Humm J L, Larson S
M and Rosenzweig K 2000 Radiotherapy treatment planning for patients with nonsmall cell lung cancer using positron emission tomography (PET) Proc. 42nd Annual
Erridge S C, Robar J L and Wu J S 2003 The use of electronic compensation with MLC in
wide field irradiation for head and neck cancers Clinical Oncol. 15 (Suppl. 2) S1314
402
References
Essers M, de Langen M, Dirkx M L P and Heijmen B J M 2001 Commissioning of a

commercially available system for intensity-modulated radiotherapy dose delivery
with dynamic multileaf collimation Radiother. Oncol. 60 21524
Esthappan J, Mutic S, Malyapa R S, Grigsby P W, Zoberi I, Dehdashti F, Miller T R,
Bosch W R and Low D A 2004 Treatment planning guidelines regarding the use of
CT/PET guided IMRT for cervical carcinoma with positive paraaortic lymph nodes
Int. J. Radiat. Oncol. Biol. Phys. 58 128997
Evans P M, Donovan E M, Partridge M, Childs P J, Convery D J, Eagle S, Hansen V N,
Suter B L and Yarnold J R 2000 The delivery of intensity modulated radiotherapy to
the breast using multiple fields Radiother. Oncol. 57 7989
Evans P M, Hansen V N and Swindell W 1997 The optimum intensities for multiple static
multileaf collimator field compensation Med. Phys. 24 114756
Evans P M, Partridge M and Symonds-Tayler J R N 2001 Sampling considerations for
intensity-modulated radiotherapy verification using electronic portal imaging Med.
Phys. 28 54352
Ezzell G A 2003 Clinical implementation of IMRT treatment planning IMRTthe State of
the Art: AAPM Medical Physics Monograph Number 29 ed J R Palta and T R Mackie
(Madison: WI: Medical Physics Publishing) pp 47593
Ezzell G and Chungbin S 2001 Low MU/segment issues with IMRT Med. Phys. 28 12534
Ezzell G A, Galvin J M, Low D, Palta J R, Rosen I, Sharpe M B, Xia P, Xiao Y, Xing L
and Yu C X 2003 Guidance document on delivery, treatment planning, and clinical
implementation of IMRT: report of the IMRT subcommittee of the AAPM radiation
therapy committee Med. Phys. 30 2089115
Ezzell G A, Schild S E and Wong W W 2000 Development of a treatment planning protocol
for prostate treatments with IMRT Proc. 42nd Annual ASTRO Meeting. Int. J. Radiat.
Oncol. Biol. Phys. 48 (Suppl.) 140
Falco T, Evans M D C, Shenouda G and Podgorsak E B 2001 Ultrasound imaging for
external beam prostate treatment setup and dosimetric verification Med. Phys. 28
1988
Faria S L, Ferrigno R and Souhami L 2002 Not treating all fields in every radiotherapy
session. Questioning an old dogma Radiother. Oncol. 64 (Suppl. 1) S192
Fayos F, Saez Beltran M and Sanchez R 2001a Measurements for implementing IMRT
with dMLC Radiother. Oncol. 61 (Suppl. 1) S107
2001b Quality assurance in IMRT with dMLC Radiother. Oncol. 61 (Suppl. 1) S48
Fenton P and Lynch R 2002 A retrospective study on the impact of image registration use
in radiation therapy treatment planning for primary brain tumour patients Radiother.
Fielding A L, Clark C H and Evans P M 2003a Verification of patient position and delivery
of IMRT by electronic portal imaging Med. Phys. 30 1352
Fielding A L and Evans P M 2001 The use of electronic portal images to verify the
delivery of intensity modulated radiotherapy beams produced with dynamic multileaf collimation Radiother. Oncol. 61 (Suppl. 1) S26
Fielding A L, Evans P M and Clark C H 2002a The use of electronic portal imaging
to verify patient position during intensity modulated radiotherapy delivered by the
dynamic MLC technique Proc. IPEM Biennial Meeting (Southampton, 1617 July)
pp 1011
2002b The use of electronic portal imaging to verify the patient set-up and delivery of
intensity modulated radiotherapy Radiother. Oncol. 64 (Suppl. 1) S12
References
403
2002c The use of electronic portal imaging to verify patient position during intensity
modulated radiotherapy delivered by the dynamic MLC technique Int. J. Radiat.
2003b A new method for verification of patient position and delivery of IMRT
using electronic portal imaging Proc. World Congress in Medical Physics (Sydney)
CDROM and website only
Fitchard E, Olivera G, Kapatoes J, Ruchala K, Reckwerdt P and Mackie R 2000
Registration of tomotherapy animal study data Proc. World Congress of Medical
Physics and the AAPM Annual Congress, August 2000 paper WE-E313-07
Fiorino C, Cozzarini C, Fellin G, Fopplano F, Menegotti L, Piazzolla A, Sanguineti G,
Vavassori V and Valdegni R 2002 Evidence of correlation between rectal DVHs and
late bleeding in patients treated for prostate cancer Radiother. Oncol. 64 (Suppl. 1)
S71
Fiorino C, Gianolini S and Nahum A E 2003 A cylindrical model of the rectum: comparing
dose-volume, dose-surface and dose-wall histograms in the radiotherapy of prostate
cancer Phys. Med. Biol. 48 260316
Fix M K, Manser P, Born E J, Mini R and Ruegsegger P 2001a Monte Carlo simulation of
dynamic MLC using a multiple source model Med. Phys. 28 1197
2001b Monte Carlo simulation of dynamic MLC using a multiple source model Phys.
Med. Biol. 46 324157
2002 Influence of dynamic MLC on photon beam energy spectra Med. Phys. 29 1270
Fogg R, Chetty I, DeMarco J, Agazaryan N and Solberg T 2000 Development and
modification of a virtual source model for Monte Carlo based IMRT verification Proc.
paper WE-B309-03
Fogliata A, Bolsi A and Cozzi L 2002 Critical appraisal of treatment techniques based on
conventional photon beams, intensity modulated photon beams and proton beams for
therapy of the intact breast Radiother. Oncol. 62 13745
2003 Comparative analysis of intensity modulation inverse planning modules of three
commercial treatment planning systems applied to head and neck tumour model
Fong P M, Keil D C, Does M D and Gore J C 2001 Polymer gels for magnetic resonance
imaging of irradiation dose distributions at normal room atmosphere Phys. Med. Biol.
46 310514
Foppiano F, Fiorino C, Frezza G, Greco C, Valdagni R and the AIRO National Working
Group on Prostate Radiotherapy 2003 The impact of contouring uncertainty on rectal
3D dose-volume data: Results of a dummy run in a multicenter trial (AIROPROS0102) Int. J. Radiat. Oncol. Biol. Phys. 57 5739
Ford E, Chang J, Mueller K, Mageras G, Yorke E, Ling C and Amols H 2002b Low-dose
megavoltage cone-beam CT with an amorphous silicon electronic portal imaging
device Med. Phys. 29 1241
Ford E C, Mageras G S, Yorke E, Rosenzweig K E, Wagman R and Ling C C 2002a
Evaluation of respiratory movement during gated radiotherapy using film and
electronic portal imaging Int. J. Radiat. Oncol. Biol. Phys. 52 52231
Forster K, Stevens C, Kitamura K, Starkschall G, Liu H, Liao Z, Chang J, Cox J, Jeter M,
Guerrero T and Komaki R 2003 3D assessment of respiration induced NSCLC tumor
motion Med. Phys. 30 1364
404
References
Fraass B 2001 Optimization and IMRT: performing valid comparisons between techniques
Fraass B A, McShan D L, Vineberg K A, Balter J, Ten Haken R K and Eisbruch 2002
Generation of robust IMRT plans for head/neck cancer: replacing the PTV with direct
accommodation of setup uncertainty and buildup region dosimetry Radiother. Oncol.
64 (Suppl. 1) S96
Francescon P, Cora S and Chiovati P 2001 Verification of an intensity modulated treatment
plan with a Monte Carlo code Radiother. Oncol. 61 (Suppl. 1) S2
Fransson A, Backfrieder W, Lorang T, Dieckmann K, Bogner J, Nowotny R and Potter R
2000a Treatment planning based on MR images Radiother. Oncol. 56 (Suppl. 1) S17
Fransson A, Bogner J, Dieckmann K and Potter R 2000b Registration of MR and
CT images using the BrainLab and Helax TMS treatment planning systemsa
comparative study Radiother. Oncol. 56 (Suppl. 1) S208
Frazier R C, Vicini F A, Sharpe M B, Fayad J, Yan D, Martinez A A Wong J W 2000
The impact of respiration on whole breast radiotherapy: A dosimetric analysis using
active breathing control Proc. 42nd Annual ASTRO Meeting. Int. J. Radiat. Oncol.
Biol. Phys. 48 (Suppl.) 200
Frazier R C, Vicini F A, Sharpe M B, Yan D, Fayad J, Baglan K L, Kestin L L,
Remouchamps V M, Martinez A A and Wong J W 2004 Impact of breathing motion
on whole breast radiotherapy: A dosimetric analysis using active breathing control
Freedman G M, Price Jnr R A, Mah D, Milestone B, Movsas B, Horwitz E, Mitra R and
Hanks G E 2001 Routine use of MRI and CT simulation for treatment planning of
intensity modulated radiation therapy (IMRT) in prostate cance Int. J. Radiat. Oncol.
Frencl L, Cernohuby P and Drnec I 2001 IMRT without MLCWhen, how and why?
Fu W, Dai J, Hu Y, Han D and Song Y 2004 Delivery time comparison for intensitymodulated radiation therapy with/without flattening filter: a planning study Phys.
Med. Biol. 49 153547
Gaboriaud G, Papatheodorou S, Pontvert D and Rosenwald J 2000 Comparison of noncoplanar three-dimensional conformal radiation therapy and non-coplanar intensitymodulated radiation therapy in the case of brain tumours Proc. World Congress of
Medical Physics and the AAPM Annual Congress, August 2000 paper MO-B309-05
Gaede S, Wong E and Rasmussen H 2001 Evaluating alternative approaches to inverse
planning optimization Med. Phys. 28 1206
2002 Evaluation of a systematic beam direction selection algorithm for IMRT Med.
Phys. 29 1335
2004 An algorithm for systematic selection of beam directions for IMRT Med. Phys.
31 37688
Gagliardi G, Lax I, Ottolenghi A and Rutqvist L E 1996 Long term cardiac mortality
after radiotherapy of beast cancerapplications of the relative seriality model Brit.
J. Radiol. 69 83946
Gall K and Chang C 2003 Accuracy of the Accuray Cyberknife dose deposition algorithm
Med. Phys. 30 1352
Gallant G and Schreiner L 2000 TOMOPlan 1: Towards a 3D RTP optimization system for
cobalt tomotherapy Proc. World Congress of Medical Physics and the AAPM Annual
Congress, August 2000 paper MO-CXH-12
References
405
Galvin J M 1999 The multileaf collimatora complete guide Proc. 1999 AAPM Annual
Meeting; see also Med. Phys. 26 10923
Galvin J, Bednarz G, Lin L, Lally B and Komarnicky L 2000a Using segmented fields
to treat the breast Proc. World Congress of Medical Physics and the AAPM Annual
Congress, August 2000 paper MO-GBR-05
Galvin J M, Chen X G and Smith R M 1993 Combining multileaf fields to modulate fluence
distributions Int. J. Radiat. Oncol. Biol. Phys. 27 697705
Galvin J M, Ezzell G, Eisbruch A, Yu C, Butler B, Xiao Y, Rosen I, Rosenman J, Sharpe
M, Xing L, Xia P, Lomax T, Low D A and Palta J 2004 Implementing IMRT in
clinical practice: a joint document of the American society for therapeutic radiology
and oncology and the American association of physicists in medicine Int. J. Radiat.
Galvin J, Xiao Y, Bednarz G and Huq S 2000b A comparison of optimized forward
planning and inverse planning for target with invaginations Radiother. Oncol. 56
(Suppl. 1) S115
Galvin J, Xiao Y, Michalski D, Censor Y, Houser C, Bednarz G and Huq M S 2001
Segmental inverse planning (SIP) that starts with a definition of allowable fields Med.
Phys. 28 1253
Garcia-Alvarez R, Liney G P and Beavis A W 2003a The use of functional MR for planning
conformal avoidance radiotherapy delivered by IMRT Clinical Oncol. 15 (Suppl. 2)
S13
2003b Use of functional magnetic resonance imaging for planning intensity
modulated radiotherapy J. Radiother. Practice 3 6672
George R, Keall P J, Kini V R, Vedam S S, Siebers J V, Wu Q, Lauterbach M H, Arthus
D W and Mohan R 2003 Quantifying the effect of intrafraction motion during breast
IMRT planning and dose delivery Med. Phys. 30 55262
Ghelmansarai F, Pouliot J, Zheng Z and Svatos M 2003 Low dose-high contrast
megavoltage cone beam CT Med. Phys. 30 1345
Gibbs I C, Sattah M, Guerrero T, Chang S D, Martin D, Kim D and Le Q T 2002 Cyberknife
image-guided radiosurgery of the spine Radiother. Oncol. 64 (Suppl. 1) S300
Gibon D, Kulik C, Poupon L, Castelain B, Vasseur C and Rousseau J 2000 Optimization
of micro-multileaf collimator irradiation parameters in conformal radiation therapy
Gierga D, Chen G, Kung J, Betke M, Lombardi J and Willett C 2003 Quantitative
fluoroscopy of respiration-induced abdominal tumor motion on IMRT dose
distributions Med. Phys. 30 1340
2004 Quantification of respiration-induced abdominal tumor motion and its impact
on IMRT dose distributions Int. J. Radiat. Oncol. Biol. Phys. 58 158495
Gierga D and Jiang S 2002 A Monte Carlo study on the interplay effect between MLC leaf
motion and respiration-induced tumor motion in lung IMRT delivery Med. Phys. 29
1266
Gillin M T 2003 Socio-economic issues of IMRT in IMRTthe State of the Art: AAPM
Medical Physics Publishing) p 82941
Giraud P, Helfre S, Servois V, Dubray B, Beigelman-Grenier C P, Zalcman-Liwartoski
G A, Straus-Zelter C M, Neuenschwander S, Rosenwald J C and Cosset J M
2000 Evaluation of intrathoracic organs mobility using CT gated by a spirometer
406
References
Glatstein E 2003a The return of the snake oil salesmen Int. J. Radiat. Oncol. Biol. Phys.
55 5612
2003b In response to Dr Amols Int. J. Radiat. Oncol. Biol. Phys. 56 15078
Glendinning A G, Hunt S G and Bonnett D E 2001 Comparison of collimator position
verification during IMRT using a commercial EPID and a novel strip ionisation
chamber Radiother. Oncol. 61 (Suppl. 1) S17
Gluckman G R, Foraci J, Meek A G and Reinstein L E 2001b A system for comprehensive
3D validation of localization, planning and treatment delivery for IMRT Int. J. Radiat.
Gluckman G, Foraci J, Meek A, Reinstein L, Radford D and Followill D 2001a Adaptation
of the RPC IMRT prostate phantom for full 3D dose verification using polymer gel
and multiple-plane radiochromic film inserts Med. Phys. 28 1269
Gluckman G and Reinstein L 2000 Evaluation of a new software package for validation of
3D-conformal treatment delivery Proc. World Congress of Medical Physics and the
AAPM Annual Congress, August 2000 paper MO-FXH-55
Goddu S, Vanek K N and Nelson S 2002 Moving target, dynamic treatment: is the
respiratory gating a solution? Med. Phys. 29 1270
Goitein M 2004 Organ and tumour motion: an overview Seminars Radiat. Oncol. 14 29
Goldman S P, Chen J Z, Battista J J 2004 Fast inverse dose optimisation (FIDO) for IMRT
via matrix inversion with no negative intensities Proc. 14th Int. Conf. on the Use of
Computers in Radiation Therapy (Seoul, May) pp 11215
Graham J D 2000 Geometrical uncertainties in prostate and bladder cancer Radiother.
Greer P B, Beckham W A and Ansbacher W 2003 Improving the resolution of dynamic
intensity modulated radiation therapy delivery by reducing the multileaf collimator
sampling distance Med. Phys 30 2793803
Greer P B and van Doorn T 2000 A design for a dual assembly multileaf collimator Med.
Phys. 27 224255
Gregoire V 2002 Multimodality image coregistration for treatment planning in HNSCC
Gregoire V, Daisne J F, Duprez T, Geets X, Hamoir M, Lonneux M and Weynand B. 2003
Pathology-based verification of head and neck imaging: how far are we from the truth
and what are the implications for 3D-CRT and IMRT? Radiother. Oncol. 68 (Suppl.
1) S16
Grigereit T, Nelms B, Dempsey J, Garcia-Ramirez J, Low D, Purdy J 2000 Compensating
filters for IMRT 1: Material characterization and process verification Proc. World
Congress of Medical Physics and the AAPM Annual Congress, August 2000 paper
TU-EBR-05
Grigorov G, Kron T, Wong E, Chen J, Sollazzo J and Rodrigues G 2003 Optimisation of
helical tomotherapy treatment plans for prostate cancer Phys. Med. Biol. 48 193343
Groll J, Bohsung J, Jahn U, Kurth C, Pfaender M, Grebe G, Budach V 2000 Dosimetric
verification of the Varian IMRT system Proc. World Congress of Medical Physics
and the AAPM Annual Congress, August 2000 paper WE-E313-03
Grosser K-H 2000 Reduction of treatment time in IMRT step and shoot irradiations by
means of a changed fraction scheme Proc. 13th Int. Conf. on the Use of Computers
in Radiation Therapy (Heidelberg, May) ed W Schlegel and T Bortfeld (Heidelberg:
Springer) pp 30810
References
407
Grosser K H, Kober B and Thilmann C 2002 Minimizing field components for delivering of
intensity-modulated radiotherapy step-and-shoot beam sequences Radiother. Oncol.
64 S213
Guan H 2002 Megavoltage portal CT using an amorphous silicon imaging device with
comparison to a CCD based EPID system Med. Phys. 29 1243
Guerrero T M, Crownover R L, Rodebaugh R F, Pawlicki T, Martin D P, Glosser G D,
Whyte R I, Le Q T, Murphy M J, Shiomi H, Weinhaus M S and Ma C 2001 Breath
holding versus real-time target tracking for respiratory motion compensation during
radiosurgery for lung tumors Int. J. Radiat. Oncol. Biol. Phys. 51 (Suppl. 1) 26
Guerrero Urbano M T and Nutting C M 2004a Clinical use of intensity-modulated
radiotherapy: part 1 Brit. J. Radiol. 77 8896
2004b Clinical use of intensity-modulated radiotherapy: part 2 Brit. J. Radiol. 77
17782
Gull S F and Skilling J 1984 The maximum entropy method in image processing IEE Proc.
Special Issue on Communications, Radar and Signal Processing 6 646
Gulliford S L, Webb S, Rowbottom C G, Corne D W and Dearnaley D P 2004 Use
of artificial networks to predict biological outcomes for patients receiving radical
radiotherapy of the prostate Radiother. Oncol. 71 312
Gum F, Bogner L, Scherer J, Bock M and Rhein B 2001 3D-dose verification of IMRT
treatment plans by means of an inhomogeneous anthropomorphic Fricke-gel phantom
Med. Phys. 28 1207
Gunawardena A, Meyer R, DSouza W, Shi L, Yang W and Naqvi S 2003 A new leafsequencing algorithm using difference matrices reduces aperture number in IMRT
Med. Phys. 30 1347
Gurgoze E and Rogers K 2002 Dosimetric replication of MLC by customized filters in the
delivery of IMRT Med. Phys. 29 1272
Gustafsson A 2000 A general algorithm for sequential multileaf modulation Proc. World
WE-E313-08
Gustavsson H, Back S A J, Lepage M, Rintoul L and Baldock C 2001a Development
and optimisation of a 2-hydroxyethylacrylate MRI polymer gel dosimeter Radiother.
Oncol. 61 S91
Gustavsson H, Back S A H and Olsson L E 2001b Verification of IMRT using gel dosimetry
Gustavsson H, Engstrom P, Haraldsson P, Magnusson P and Nystrom H 2000 3D
dosimetric verification of IMRT using polymer gel and MRI Radiother. Oncol. 56
(Suppl. 1) S205
Gustavsson H, Haraldsson P, Karksson A, Engstrom P, Back S A J, Nystrom H and Olsson
L E 2002 IMRT verification using magic-type polymer gel Radiother. Oncol. 64
(Suppl. 1) S43
Ha S W 2004 Physicians perspective of use of computers in RT Proc. 14th Int. Conf. on
the Use of Computers in Radiation Therapy (Seoul, May) pp 13
Hagekyriakou J, Song G L and Sephton R G 2000 The accuracy of an ultrasound guidance
system for positioning the prostate at radiotherapy Radiother. Oncol. 56 (Suppl. 1)
S38
Hajdok G, Kerr A and Schreiner L 2002 An investigation of megavoltage computed
tomography using a Cobalt-60 gamma ray source for radiotherapy treatment
verification Med. Phys. 29 1341
408
References
Halperin E C 2000 Overpriced technology in radiation oncology Int. J. Radiat. Oncol. Biol.
Phys. 48 91718
Hansen V N, Evans P M, Budgell G J, Mott J H L, Williams P C, Brugmans M J P,
Wittkamper F W, Mijnheer B J and Brown K 1998 Quality assurance of the dose
delivered by small radiation segments. Phys. Med. Biol. 43 266575
Happersett L, Hunt M, Chong L, Chui C, Spirou S, Burman C and Amols H 2000 Intensity
modulated radiation therapy for the treatment of thyroid cancer Proc. 42nd Annual
ASTRO Meeting. Int. J. Radiat. Oncol. Biol. Phys. 48 351
Happersett L, Mageras G S, Zelefsky M J, Burman C M, Leibel S A, Chui C, Fuks Z, Bull
S, Ling C and Kutcher G J 2003 A study of the effects of internal organ motion on
dose escalalation in conformal prostate treatments Radiother. and Oncol. 66 26370
Hardemark B, Rehbinder H and Lof J 2003 Rotating the MLC between segments improves
performance in step-and-shoot IMRT delivery Med. Phys. 30 1405
Harnisch G A, Willoughby T R and Kupelian P A 2000 Irradiation of dominant
intraprostatic lesions with short-course intensity modulated radiotherapy: Dosimetry
and planning Proc. 42nd Annual ASTRO Meeting. Int. J. Radiat. Oncol. Biol. Phys.
48 (Suppl.) 205
Hartkens T, Rueckert D, Schnabel J A, Hawkes D J and Hill D L G 2002 VTK CISG
Registration Toolkit: an open source software for affine and non-rigid registration of
single and multiple 3D images BVM 2002 (Leipzig: Springer)
Hartmann M, Bogner L and Fippel M 2002a IMRT: Inaccuracies of a conventional
inverse pencil beam TPS, obtained by a comparison with a Monte-Carlo based TPS
Hartmann M, Bogner L, Scherer J and Scherer S 2001 IMRT optimisation based on a new
inverse Monte-Carlo code Radiother. Oncol. 61 (Suppl. 1) S47
Hartmann G H and Fohlisch F 2002 Dosimetric characterization of a new miniature
multileaf collimator Phys. Med. Biol. 47 N1717
Hartmann G, Pastyr O, Echner G, Hadinger U, Seeber S, Juschka J, Richter J and Schlegel
W 2002b A new design for a mid-sized multileaf collimator (MLC) Radiother. Oncol.
64 (Suppl. 1) S214
Hawliczek R, Nespor W, Schmidt W and Pawlas K 2000 IMRT at an European community
hospitalone year of experience Radiother. Oncol. 56 (Suppl. 1) S196
Hector C L 2001 The effect of patient motion on the delivery of intensity-modulated
radiotherapy PhD Thesis London University
Heijmen B J M 2000 New applications for your electronic portal imaging device (EPID)
Henrys A, Taylor H, Bedford J and Tait D 2003 Optimizing the use of multileaf collimators
in the treatment of colo-rectal cancer Clinical Oncol. 15 (Suppl. 2) S18
Heon J, Falco T, Shenouda G, Duclos M, Faria S and Souhami L 2002 Ultrasound imaging
for external beam prostate treatment setup Radiother. Oncol. 64 (Suppl. 1) S45
Herman M G, Rosenzweig D P and Hendee W R 2003 Every patient receiving 3D or IMRT
must have image-based target localization prior to turning on the beam Med. Phys.
30 2879
Higgins P D, Alaei P, Gerbi B J and Dusenbery K E 2003 In vivo diode dosimetry for
routine quality assurance in IMRT Med. Phys. 30 311823
Hills M, Audet C, Duzenli C and Jirasek A 2000a Polymer gel dosimetry using x-ray
computed tomography: a feasibility study Phys. Med. Biol. 45 255971
References
409
Hills M, Jirasek A, Audet C and Duzenli C 2000b X-ray CT polymer gel dosimetry:
applications to stereotactic radiosurgery and proton therapy Radiother. Oncol. 56
(Suppl. 1) S80
Hoban P 2000 Conversion of leaf sequences to dose distributions in micro-IMRT Proc.
paper MO-BBR-03
2002 Use of a mini-multileaf collimator for simultaneous IMRT boost in prostate
cancer treatment Med. Phys. 29 1268
Hoban P, Short R, Biggs D, Rose A, Smee R and Schneider M 2001 Clinical use of a
micro-IMRT system Med. Phys. 28 12023
Hover K 2003 A quantitative comparison on characteristics of primary-, mini-and micro
multileaf collimators Proc. World Congress in Medical Physics (Sydney) CDROM
and website only
Hover K H, Haering P, Jahn U and Krohnholz H L 2001 Dosimetric characteristics of
different MLCs an intercomparison Radiother. Oncol. 61 (Suppl. 1) S83
Hover K H, Haering P, Krohnholz H, Wurm R and Jahn U 2000 A dosimetrical
intercomparison of different micro multileaf collimators Radiother. Oncol. 56 (Suppl.
1) S97
Hofman P P, Nederveen A A J, van der Heide U U A, van Moorselaar J R J A, Dehnad H
H and Lagendijk J J J 2002 Rectal dose in intensity modulated radiotherapy of the
prostate: comparison with the previous conformal irradiation technique Radiother.
Holloway L C and Hoban P W 2001 Combining a physically based optimization method
with a biologically based tissue importance weight changing method Int. J. Radiat.
Oncol. Biol. Phys. 51 (Suppl. 1) 72
2002 Iterative optimization of beam positions for IMRT Med. Phys. 29 1257
Holmberg O, Kasenter A, Verdugo V, Buckney S, Gribben A, Hollywood D and
McClean B 2000 Temporal uncertainties of target volumes in step-and-shoot IMRT:
a numerical simulation Radiother. Oncol. 56 (Suppl. 1) S108
Holmes T W 2001 A method to incorporate leakage and head scatter corrections into a
tomotherapy inverse treatment planning algorithm Phys. Med. Biol. 46 1127
Holmes T W, Shepard D, Naqvi S, Li A, Ma L and Yu C X 2000 Improvements to pencil
beam-based inverse treatment planning methodology Proc. 42nd Annual ASTRO
Hong L H, Alektiar K, Chui C, LoSasso T, Hunt M, Spirou S, Yang J, Amols H, Ling C,
Fuks Z and Leibel S 2002 IMRT of large fields: Whole-abdomen irradition. Int. J.
Radiat. Oncol. Biol. Phys. 54 27889
Hoogeman M S, Koper P C M, Heemsbergen W D, van Putten W, Boersma L J, Jansen
P P, Levendag P C and Lebesque J V 2002 Correlation of 3D dose distribution and
rectal complications after prostate treatment Radiother. Oncol. 64 (Suppl. 1) S71
Hoogeman M S, van Herk M, de Bois J, Boersma L J, Koper P C M and Lebesque J V
2003 Strategies to reduce the systematic error due to tumor motion in radiotherapy of
prostate cancer Phys. Medica. 19 545
Hoogeman M S, van Herk M, de Bois J, Muller-Timmermans P, Koper P C M and Lebesque
J V 2004 Quantification of local rectal wall displacements by virtual rectum unfolding
Horn B K P and Schunk B G 1981 Determining optical flow Artificial Intelligence 17
185204
410
References
Hossain M, Houser C and Galvin J 2001 The effect of switch rates on the dose output from
an intensity modulating dynamic multileaf collimator Med. Phys. 28 1283
Hou Q and Wang Y 2001 Molecular dynamics used in radiation therapy Phys. Rev. Lett.
87 16810114
Hou Q, Wang J, Chen Y and Galvin M 2003a An optimisation algorithm for intensity
modulated radiotherapyThe simulated dynamics with dose-volume constraints
Med. Phys. 30 618
2003b Beam orientation optimization for IMRT by a hybrid method of the genetic
algorithm and the simulated dynamics Med. Phys. 30 23607
Hou Q, Zhang C, Wu Z and Chen Y 2004 A method to improve spatial resolution and
smoothness of intensity profiles in IMRT treatment planning Med. Phys. 31 133947
Hough J and Jones D 2002 Robotic patient positioning for radiotherapy Med. Phys. 29
1213
Hounsell A R, Mott J H L, Budgell G J and Wilkinson J M 2001 Dose model considerations
for IMRT treatments using multileaf collimators Phys. Medica. 17 161
Hsiung C-Y, Yorke E D, Chui C-S, Hunt M A, Ling C C, Huang E-Y, Wang C-J, Chen
H-C, Yeh S-A, Hsu H-C and Amols H I. 2002 Intensity-modulated radiotherapy
versus conventional three-dimensional conformal radiotherapy for boost or salvage
treatment of nasopharyngeal carcinoma Int. J. Radiat. Oncol. Biol. Phys. 53 63847
Hu N, Downey D B, Fenster A and Ladak H M 2003 Prostate boundary segmentation from
3D ultrasound images Med. Phys. 30 164859
Huang E, Dong L, Chandra A, Kuban D A, Rosen I I, Evans A and Pollack A 2002a
Intrafraction prostate motion during IMRT for prostate cancer Int. J. Radiat. Oncol.
Biol. Phys. 53 2618
Huang E, Dong L, Chandra A, Kuban D A, Rosen I I and Pollack A 2001b Intrafraction
prostate motion during IMRT for prostate cancer Int. J. Radiat. Oncol. Biol. Phys. 51
21213
Huang J, Katz A, Haas J, Urrutia T, Lauritano J and MacMelville W 2002b IMRT treatment
of prostate cancer using 6MV photon beam Med. Phys. 29 1254
Huang J, Katz A, Haas J, Urrutia T, Lauritano J and Ortiz E 2001a Using beak slit with
MIMiC in clinical IMRT Med. Phys. 28 1202
Huang J, Urrutia T, Lauritano J and Katz A 2000 Application of gafchromic film in clinical
QA of IMRT Proc. World Congress of Medical Physics and the AAPM Annual
Huang T-C, Guerrero T, Zhang G, Lin H-D, Kuban D and Lin K-P 2004 3D optical flow
method: Speed, accuracy, and convergence in radiotherapy applications Proc. 14th
Int. Conf. on the Use of Computers in Radiation Therapy (Seoul, May) pp 56973
Hugo G D, Agazaryan N and Solberg T D 2002 An evaluation of gating window size,
delivery method, and composite field dosimetry of respiratory-gated IMRT Med.
Phys. 29 251725
2003a The effects of target motion on planning and delivery of respiratory-gated
IMRT Med. Phys. 30 1330
Hugo G, Solberg T, Demarco J, Paul T, Agazaryan N, Arellano A and Smathers J 2001
Preliminary dosimetric investigations into gated IMRT Med. Phys. 28 1294
Hugo G, Tenn S and Agazaryan N 2003b An evaluation of intrafraction motion-induced
error for fractionated IMRT delivery Med. Phys. 30 1470
Huq M S, Das I J, Steinberg T and Galvin J M 2002 A dosimetric comparison of various
multileaf collimators Phys. Med. Biol. 47 N15970
References
411
Humm J, Erdi Y, Nehmeh S, Pugachev A, Zanzonico P and Ling C 2003 The potential role
of PET in radiation treatment planning Radiother. Oncol. 68 (Suppl. 1) S28
Hunt M A, Hsiung C-Y, Spirou S V, Chui C-S and Amols H I 2000 Evaluation of concave
dose distributions created using an inverse planning system Proc. 42nd Annual
Hunt M A, Hsiung C-Y, Spirou S V, Chui C-S, Amols H I and Ling C C 2002 Evaluation
of concave dose distributions created using an inverse planning system Int. J. Radiat.
Huntzinger C, and Hunt M 2000 A simple method to estimate IMRT monitor units Proc.
paper MO-BBR-07
Hurkmans C W, Cho B C J, Damen E, Zijp L and Mijnheer B J 2002 Reduction of
cardiac and lung complication probabilities after breast irradiation using conformal
radiotherapy with and without intensity modulation Radiother. Oncol. 62 16371
Hutchison P and Halperin E C 2002 The hidden persuaders: subtle advertising in radiation
oncology Int. J. Radiat. Oncol. Biol. Phys. 54 98991
Hwang A, Verhey L and Xia P 2001 Using a leaf sequencing algorithm to enlarge treatment
field length in IMRT Med. Phys. 28 1261
Intensity-modulated Radiation Therapy Collaborative Working Group 2001 Intensitymodulated radiotherapy: current status and issues of interest Int. J. Radiat. Oncol.
Biol. Phys. 51 880914
Iori M, Paiusco M, Nahum A, Marini P, Iotti C, Polico R, Venturi A, Armaroli L and Borasi
G 2003 IMRT: the Reggio Emilia experience Radiother. Oncol. 68 (Suppl. 1) S62
Ipe N, Roesier S, Jiang S and Ma C 2000 Neutron measurements for intensity modulated
radiation therapy Proc. World Congress of Medical Physics and the AAPM Annual
Congress, August 2000 paper FR-EA328-01
Islam M, Ramaseshan R, Norrlinger B. Gutierrez, Alasti H and Heaton R 2001 Clinical
implementation of a commercial IMRT system Med. Phys. 28 1207
Iyadurai R, Subramanium B, Rao P, John S and Ayyangar K 2003 A comparison of
manual MLC for telecobalt unit with linac-based MLC in conformal radiotherapy and
intensity modulated radiotherapy Proc. World Congress in Medical Physics (Sydney)
Jackson A 2002 Dose-volume relationships in rectal bleeding after conformal radiotherapy
for prostate cancer Radiother. Oncol. 64 (Suppl. 1) S10
Jaffray D 2002 Flat-panel cone-beam CT: an emerging technology of image-guided
radiation therapy Radiother. Oncol. 64 (Suppl. 1) S75
2003 X-ray guided IMRT IMRTthe State of the Art: AAPM Medical Physics
Publishing) pp 70326
Jaffray D A, Siewerdsen J H, Wong J W and Martinez A A 2002 Flat-panel cone-beam
computed tomography for image-guided radiation therapy. Int. J. Radiat. Oncol. Biol.
Phys. 53 133749
James H, Atherton S, Budgell G, Kirby M and Williams P 2000 Verification of dynamic
multileaf collimation using an electronic portal imaging device Phys. Med. Biol. 45
495509
James H V, Crosbie J C, Poynter A J, Bulusu V R, LeVay J, Morgan J S, Scrase C D,
and Hardy V 2002a IMRT in routine clinical usepracticalities within a small, nonacademic hospital Proc. IPEM Biennial meeting (Southampton, 1617 July) p 35
412
References
James H V, Poynter A J, Crosbie J C, MacKenzie L C, Boston S L and LeVay J 2002b

Electronic compensation for CT planned breast treatments Radiother. Oncol. 64
(Suppl. 1) S133-S-4
James H V, Poynter A J, Crosbie J, Smith S, Le Vay J, Bulusu V R, Morgan J, Scrase C D
and Hardy V 2003 Establishing IMRT in a small oncology centre Clinical Oncol. 15
(Suppl. 2) S6
James H V, Scrase C D and Poynter A J 2004 Practical experience with intensity-modulated
radiotherapy Brit. J. Radiol 77 314
Jayaraman S 2003 IMRT, 3D-CRT, inverse planning, etc., are new buzz words, but
correspond to the same old historical concepts Proc. World Congress in Medical
Physics (Sydney) CDROM and website only
Jeraj R, Balog J, Wenman D, Olivera G, Pearson D, Reckwerdt P and Mackie T R 2001
Monte Carlo simulation of the tomotherapy prototype Radiother. Oncol. 61 (Suppl.
1) S27
Jeraj R and Keall P 2000 The effect of statistical uncertainty on inverse treatment planning
based on Monte Carlo dose calculation Phys. Med. Biol. 45 360113
Jeraj R, Keall P J and Siebers J V 2002a The effect of dose calculation accuracy on inverse
treatment planning Phys. Med. Biol. 47 391407
Jeraj R, Mackie T R, Balog J, Olivera G, Pearson D, Kapatoes J, Ruchala K and Reckwerdt
P 2004 Radiation characteristics of helical tomotherapy Med. Phys. 31 396404
Jeraj R, Wu C and Mackie T R 2002b Clinical importance of optimisation convergence and
local minima in inverse treatment planning Radiother. Oncol. 64 (Suppl. 1) S97
Jiang S 2004 Tracking tumor with dynamic MLC: Be SMART Proc. 14th Int. Conf. on the
Use of Computers in Radiation Therapy (Seoul, May) p 47
Jiang S B, Bortfeld T, Trofimov A, Rietzel E, Sharp G, Choi N and Chen G T Y 2003a
Synchronized moving aperture radiation therapy (SMART): Plan optimization and
MLC leaf sequencing based on 4D CT data Med. Phys. 30 1384
2004a Synchronized moving aperture radiation therapy (SMART): Treatment
planning using 4D CT data Proc. 14th Int. Conf. on the Use of Computers in
Jiang S B and Doppke K 2001b Dosimetric effect of respiratory motion on the treatment
of breast cancer with tangential fields Med. Phys. 28 (6), 1228
Jiang S B, Kung J H, Zygmanski P and Chen G T Y 2001a A method to control the spatial
distribution of target dose heterogeneities in IMRT inverse planning Med. Phys. 28
1205
Jiang S B, Neicu T and Zygmanski P 2003b Synchronized moving aperture radiation
therapy (SMART): Superimposing tumor motion on IMRT MLC leaf sequences
under realistic delivery limitations Med. Phys. 30 1399
Jiang S B, Pope C, Al Jarrah K M, Kung J H, Bortfeld T and Chen G T Y 2003c An
experimental investigation on intra-fractional organ motion effects in lung IMRT
treatments Phys. Med. Biol. 48 177384
Jiang S, Rivinius C, Al Jarrah K, Kung J, Bortfeld T and Chen G 2002 An experimental
investigation on interplay effect between DMLC leaf motion and respiration-induced
tumor motion in lung IMRT treatment Med. Phys. 29 1347
Jiang Z, Shepard D M, Earl M A and Yu C X 2004b Asymptotic limit of few field IMRT
pp 22831
References
413
Johansen S, Eilertsen K and Furre T 2003 Comparison of treatment plans using

conventional RT and IMRT for the treatment of 10 patients with different diagnosis
Jolly J, Ding M, Pawlicki T, Chu J and Ma C-M 2001 Monte Carlo simulation of the effect
of thoracic motion on breast treatment Med. Phys. 28 1228
Jones A O, Das I J and Jones F L 2003 A Monte Carlo study of IMRT beamlets in
inhomogeneous media Med. Phys. 30 296300
Jones L and Hoban P 2000 The effect of movement on optimised IMRT dose distributions
Proc. World Congress of Medical Physics and the AAPM Annual Congress, August
2000 paper MO-CXH-03
2002 A comparison of physically and radiobiologically based optimisation for IMRT
Med. Phys. 29 144755
Ju S G, Ahn Y C, Huh S J and Yeo I J 2002 Film dosimetry for intensity modulated
radiation therapy: Dosimetric evaluation. Med. Phys. 29 3515
Jursinic P, Zhu R and Nelms B 2002 An implementation of solid filter (compensator)-based
intensity modulated radiation therapy for clinical use Med. Phys. 29 1266
Kagan A R, Schulz R J 2003 A commentary on dose escalation and bned in prostate cancer
Kamath S, Sahni S, Li J, Palta J and Ranka S 2003 Leaf sequencing algorithms for
segmented multileaf collimation Phys. Med. Biol. 48 30724
Kamath S, Sahni S, Palta J and Ranka S 2004a Algorithms for optimal sequencing of
dynamic multileaf collimators Phys. Med. Biol. 49 3354
Kamath S, Sahni S, Palta J, Ranka S and Li J 2004b Optimal leaf sequencing with
elimination of tongue-and-groove underdosage Phys. Med. Biol. 49 N719
Kapatoes J M, Olivera G H, Balog J P, Keller H, Reckwerdt P J and Mackie T R 2001a
On the accuracy and effectiveness of dose reconstruction for tomotherapy Phys. Med.
Biol. 46 94366
Kapatoes J M, Olivera G H, Lu W, Reckwerdt P J, Keller H, Balog J and Mackie T R 2000
Dose gradients as a tool in the optimization and verification of intensity modulated
radiation therapy (IMRT) Proc. 42nd Annual ASTRO Meeting. Int. J. Radiat. Oncol.
Biol. Phys. 48 139
Kapatoes J, Olivera G, Reckwerdt P, Ruchala K, Jeraj R and Mackie R 2001b A strategy
for rapidly adjusting dose distributions Radiother. Oncol. 61 (Suppl. 1) S16
Kapatoes J M, Olivera G H, Ruchala K J and Mackie T R 2001c On the verification of the
incident energy fluence in tomotherapy IMRT Phys. Med. Biol. 46 295365
Kapatoes J M, Olivera G H, Ruchala K J, Smilowitz J B, Reckwerdt P J and Mackie T R
2001d A feasible method for clinical delivery verification and dose reconstruction in
tomotherapy Med. Phys. 28 52842
Kapatoes J, Ruchala K, Olivera G, Welsh J, Forrest L, Turek M, Mackie R, Jaradat
H, Jeraj R and Reckwerdt P 2003 Internal 3D patient dose verification in imageguided radiotherapy Proc. World Congress in Medical Physics (Sydney) CDROM
and website only
Kapulsky A, Mullokandov E, Gejerman G, Saini A, Rebo I and Cardel F 2001 An
automated phantom-film QA procedure for IMRT treatments Med. Phys. 28 1284
Kavanagh B D 2003 The emperors new isodose curves Med. Phys. 30 255960
Keall P 2004 4-dimensional computed tomography imaging and treatment planning
Seminars Radiat. Oncol. 14 8190
414
References
Keall P, Kini V, Vedam S and Mohan R 2001b 4D IMRT: Dosimetric and practical
considerations Med. Phys. 28 1252
Keall P J, Siebers J V, Arnfield M, Lim J O and Mohan R 2001a Monte Carlo dose
calculation for dynamic IMRT treatments Phys. Med. Biol. 46 92941
Keall P, Siebers J, Kim J and Mohan R 2002b The clinical implementation of Monte Carlo
for IMRT Radiother. Oncol. 64 (Suppl. 1) S105
Keall P J, Starkshall G, Shukla H, Forster K M, Ortiz V, Stevens C W, Vedam S S, George
R, Guerrero T and Mohan R 2004 Acquiring 4D thoracic CT scans using a multislice
helical method Phys. Med. Biol. 49 205367
Keall P, Vedam S, George R, Kinl V and Mohan R 2002a 4D IMRT: a phantom study
Keall P J and Williamson J F 2003 Clinical evidence that more precisely defined dose
distributions will improve cancer survival and decrease morbidity Med. Phys. 30
12812
Keall P, Wu Q, Wu Y and Kim J O 2003 Dynamic MLC IMRT IMRTthe State of the
Art: AAPM Medical Physics Monograph Number 29 ed J R Palta and T R Mackie
Keane J T, Fontenla D P and Chui C S 2000 Applications of IMAT to total body radiation
(TBI) Proc. 42nd Annual ASTRO Meeting. Int. J. Radiat. Oncol. Biol. Phys. 48
(Suppl.) 239
Keller H, Fix M K, Liistro L and Ruegsegger P 2000 Theoretical considerations to the
verification of dynamic multileaf collimated fields with a SLIC-type portal image
detector Phys. Med. Biol. 45 253145
Keller-Reichenbecher M A, Bortfeld T, Levegrun S, Stein J, Preiser K and Schlegel W
1999 Intensity modulation with the step and shoot technique using a commercial
MLC: a planning study Int. J. Radiat. Oncol. Biol. Phys. 45 131524
Kerambrun A, Chauvet I, Parent L, Rosenwald J, Torzsok O, Mazal A, Zefkili S,
Drouard J, Gaboriaud G and Mijnheer B 2003 Dynamic multileaf collimator (DMLC)
compensation for breast treatments: use of an anthropomorphic phantom to check the
validity of dose computation and assess the improvement of dose distribution Proc.
Kermode R H and Lawrence G P 2001 Validation of the use of data from a treatment
planning system in the design of simple lead compensators Radiother. Oncol. 61
(Suppl. 1) S100
Kermode R H, Lawrence G P and Lambert G D 2001a The use of compensators to improve
dose distributions in the head and neck region Radiother. Oncol. 61 (Suppl. 1) S100
Kermode R H, Richmond N D and Lambert G D 2001b Dosimetric verification of multisegmented treatment fields planned on Helax-TM Phys. Medica. 17 164
Kerr A T, Salomons G J and Schreiner J 2001 Dose delivery accuracy of a scanned pencil
beam for cobalt-60 tomotherapy studies Med. Phys. 28 1986
Kessler M, Ten Haken R, Pickett B and Cao Y 2003 Use of MRI and PET data for CT-based
treatment planning Radiother. Oncol. 68 (Suppl. 1) S28
Kestin L L, Sharpe M B, Frazier R C, Vicini F A, Yan D, Matter R C, Martinez A A and
Wong J W 2000a Intensity modulation to improve dose uniformity with tangential
breast radiotherapy: initial clinical experience Int. J. Radiat. Oncol. Biol. Phys. 48
155968
References
415
2000b Intensity modulation to improve dose uniformity with tangential breast

radiotherapy: initial clinical experience Int. J. Radiat. Oncol. Biol. Phys. 48 (Suppl.
3) 295
Kim D J W, Murray B R, Halperin R and Roa W H Y 2001 Held-breath self-gating
technique for radiotherapy of non-small-cell lung cancer: A feasibility study Int.
Kim J, Siebers J and Keall P 2002 Dosimetric verification of IMRT fields using an
amorphous silicon flat panel imager and Monte Carlo simulation Med. Phys. 29 1197
Kim K-H, Oh Y-K, Cho M-J, Kim J-K, Shin K-C, Kim J-K and Jeong D-H 2004c The
preliminary results of intensity-modulated radiation therapy using liquid shielding
pp 7347
Kim S, Akpati H, Dempsey J and Palta J 2003 Time-integrated dose volume histogram
(TDVH), a new paradigm for plan evaluation Med. Phys. 30 1347
Kim S, Akpati H, Dempsey J, Palta J and Ye S-J 2004a Accounting for target localization
uncertainty in DVHs: is patient-specific displacement data needed? Proc. 14th Int.
Kim S, Akpati H C, Kielbasa J E, Li J G, Liu C, Amdur R J and Palta J R 2004b Evaluation
of intrafraction patient movement for CNS and head and neck IMRT Med. Phys. 31
5006
Kim S and Palta J 2003 Multi-source intensity-modulated radiation beam delivery system
and method US Patent 6,449,336
Kim S, Yi B, Choi E and Ha S 2003 Impact of CT scan time on reproducibility of
respiratory movement at the time of radiotherapy for lung cancer Proc. World
Congress in Medical Physics (Sydney) CDROM and website only
King G, Selvaraj R, Mogus R and Wu A 2002 Comparison of beam on time for sliding
window and step and shoot IMRT treatment plans Med. Phys. 29 1271
Kini V R, Keall P J, Vedam S S, Arthur D W, Kavanagh B D, Cardinale R M and
Mohan R 2000 Preliminary results from a study of a respiratory motion tracking
system: Underestimation of target volume with conventional CT simulation Proc.
42nd Annual ASTRO Meeting. Int. J. Radiat. Oncol. Biol. Phys. 48 (Suppl.) 164
Kini V R, Vedam S S, Keall P J and Mohan A R 2001 A dynamic non-invasive technique
for predicting organ motion in respiratory-gated radiotherapy of the chest Int. J.
Radiat. Oncol. Biol. Phys. 51 (Suppl. 1) 25
Kirby M C A and Shentall G S 2001 Why standard treatment margins may need to be
modified for IMRT techniques Radiother. Oncol. 61 (Suppl. 1) S62
Kirby M C, Shentall G S, Bulmer N C, Glendinning A G and McKay D M 2002 Verifying
step and shoot IMRT fluence maps using Elekta iView and iViewGT electronic portal
imaging devices (EPIDs) Proc. IPEM Biennial Meeting (Southampton, 1617 July)
pp 1314
Kitamura K, Shirato H and Seppenwoolde Y 2001 3D intra-fractional movement of
prostate measured during real-time tumor-tracking radiation therapy (RTRT) in
supine and prone treatment positions Int. J. Radiat. Oncol. Biol. Phys. 51 (Suppl.
1) 213
Kitamura K, Shirato H, Seppenwoolde Y, Onimaru R, Oda M, Fujita K, Shimizu S,
Shinohara N, Harabayashi T and Miyasaka K 2002 Three-dimensional intrafractional
movement of prostate measured during real-time tumor-tracking radiotherapy in
supine and prone treatment positions Int. J Radiol Oncol. Biol. Phys. 53 111723
416
References
Klabbers B M, de Munck J C, Sloman B J, Langendijk J A, deBree R, Hoekstra O S,

Boellaard R and Lammertsma AA 2002 Matching PET and CT scans of the head and
neck area: development of method and validation Radiother. Oncol. 64 (Suppl. 1)
S209
Klein E E and Low D A 2000 Interleaf leakage testing for 0.5 and 1.0 cm DMLC systems
incorporating patient motion Proc. 42nd Annual ASTRO Meeting. Int. J. Radiat.
2001 Interleaf leakage for 5 and 10 mm dynamic multileaf collimating systems
incorporating patient motion Med. Phys. 28 170310
Klein E E, Low D A, Sohn J W and Purdy J A 2000 Differential dosing of prostate and
seminal vesicles using dynamic multileaf collimation Int. J. Radiat. Oncol. Biol.
Phys. 48 144756
Kneschaurek P, Geinitz H and Zimmermann F 2000 Daily position corrected dose
histograms of the rectum Proc. World Congress of Medical Physics and the AAPM
Annual Congress, August 2000 paper MO-FXH-50
Knoos T, Karolak L and Nilsson P 2001 Quantification of dosimetry changes in a TPS after
implementation of a new dose calculation algorithm: pencil beam vs collapsed cone
Koelbl O, Bratengeier K and Flentje M 2002 Intensity modulated radiotherapy (IMRT)
versus 3-D conformal, conventional radiotherapy (3D-RT) in lung cancer Radiother.
Oncol. 64 S259
Komanduri K 2002 Dosimetric considerations for breast cancer treatment using intensity
modulated radiotherapy Med. Phys. 29 1273
Korevaar E W, Huizenga H, Lof, Stroom J C, Leer J W H and Brahme A 2002 Investigation
of the added value of high-energy electrons in intensity-modulated radiotherapy:
Four clinical cases Int. J. Radiat. Oncol. Biol. Phys. 52 23653
Korreman S S, Pedersen A N, Specht L and Nystrom H 2003 Breathing adapted
radiotherapy (BART) for breast cancer Radiother. Oncol. 68 (Suppl. 1) S40
Kry S, Salehpour M, Followill D, Stovall M and Rosen I 2003 Risk assessment of
secondary malignancies from IMRT treatments Med. Phys. 30 1330
Kron T, Battista J, Bauman G and van-Dyk J 2004a On the role of helical tomotherapy in
clinical practice: Results from planning studies Proc. 14th Int. Conf. on the Use of
Kron T, Chen J, Wong E, VanDyk J, Rodrigues G, Bauman G, Dar R, Yu E and Battista
J 2002 Initial experience with tomotherapy treatment planning in London, Ontario
Kron T, Wong E, Yartsev S, Grigorov G, Rodrigues G, Yu E, Chen J and Van Dyk J 2004b
Helical tomotherapy planning: Overlap between target and critical structures as a
predictor of plan quality Proc. 14th Int. Conf. on the Use of Computers in Radiation
Kubo H D and Wang L 2000 Compatibility of Varian 2100C gated operations with
enhanced dynamic wedge and IMRT dose delivery Med. Phys. 27 17328
Kuefer K-H, Monz M, Scherrer A, Alonso F, Trinkaus H, Bortfeld T and Thieke C
2003 Real-time inverse planning using a precomputed multi-criteria plan database
Kulidzhanov F, Sabbas A, Trichter S, Wang J and Nori D 2003 Monitor unit linearity for
the IMRT step and shoot technique Med. Phys. 30 1494
References
417
Kung J H and Chen G T Y 2000a Intensity modulated radiotherapy dose delivery error
from radiation field offset inaccuracy Med. Phys. 27 161722
2000b Monitor unit verification calculation in IMRT as a patient specific dosimetry
QA. Proc. 13th Int. Conf. on the Use of Computers in Radiation Therapy (Heidelberg,
May) ed W Schlegel and T Bortfeld (Heidelberg: Springer) pp 2923
Kung J H, Chen G T Y and Kuchnir F K, 2000b A monitor unit verification calculation
in intensity modulated radiotherapy as a dosimetry quality assurance Med. Phys. 27
222630
Kung J, Reft C, Jackson W and Abdalla I 2000a Intensity modulated radiotherapy for a
prostate patient with a prosthetic hip Proc. World Congress of Medical Physics and
the AAPM Annual Congress, August 2000 paper TU-GBR-04
Kung J H, Zygmanski P, Choi N and Chen G T Y 2003 A method of calculating a lung
clinical target volume DVH for IMRT with intrafractional motion Med Phys. 30
11039
Kung J H, Zygmanski P, Kooy H and Chen G 2000c Dose error in IMRT treatment of lung
cancer without respiratory gating Proc. 42nd Annual ASTRO Meeting. Int. J. Radiat.
Kupelian P A, Reddy C A, Klein E A and Willoughby T R 2001 Short-course intensity
modulated radiotherapy (70 Gy at 2.5 Gy per fraction) for localized prostate cancer:
Late toxicity and quality of life Int. J. Radiat. Oncol. Biol. Phys. 51 (Suppl. 1) 112
Kuriyama K, Onishi H, Sano N, Komiyama T, Aikawa Y, Tateda Y, Araki T and Uematsu
M 2003 A new irradiation unit constructed of self-moving gantry-CT and linac Int.
Kuterdem H G and Cho P S 2001 Leaf sequencing with secondary beam blocking under
leaf positioning constraints for continuously modulated radiotherapy beams Med.
Phys. 28 894902
Kwong D L, Pow E, McMillan A, Sham J and Au G 2003 Intensity-modulated radiotherapy
for early stage nasopharyngeal carcinoma: preliminary results on parotid sparing Int
J. Radiat. Oncol. Biol. Phys. 57 (Suppl.) S303
Lagendijk J J W and Bakker C J G 2000 MRI guided radiotherapy: a MRI based linear
accelerator Radiother. Oncol. 56 (Suppl. 1) S60
Lagendijk J J W, Raaymakers B W, van der Heide U A, Topolnjak R, Dehnad H, Hofman
P, Nederveen A J, Schulz I M, Welleweerd J and Bakker C J G 2002 MRI guided
radiotherapy: MRI as position verification system for IMRT Radiother. Oncol. 64
(Suppl. 1) S756
Landau D, Adams E J, Webb S and Ross G 2001 Cardiac avoidance in breast radiotherapy:
a comparison of simple shielding techniques with intensity modulated radiotherapy
Langen K M and Jones D T L 2001 Organ motion and its management. Int. J. Radiat.
Langen K M, Pouliot J, Anezinos C, Aubin M, Gottschalk A R, Hsu I-C, Lowther D, Liu
Y-M, Shinohara K, Verhey L J, Weinberg V and Roach M III 2004 Evaluation of
ultrasound-based localization for image-guided radiotherapy Int. J. Radiat. Oncol.
Langer M 2000 Application of coloring theory to reduce intensity modulated radiotherapy
dose calculations Med. Phys. 27 207783
418
References
2003 What is different about IMRT? IMRTthe State of the Art: AAPM Medical
Physics Publishing) pp 199-219
Langer M P, Thai V and Papiez L 2000 Achievable reductions in beam time and segment
number with IMRT leaf sequences Proc. 42nd Annual ASTRO Meeting. Int. J. Radiat.
2001a Comparison of leaf sequencing algorithms against the efficient frontier of
monitor unit and segment usage Med. Phys. 28 1307
2001b Improved leaf sequencing reduces segments or monitor units needed to deliver
IMRT using multileaf collimators Med. Phys. 28 24508
2001c Tradeoffs between segments and monitor units are not required for static field
IMRT delivery Int. J. Radiat. Oncol. Biol. Phys. 51 (Suppl. 1) 75
Large C, Brumley K and Ramsey C 2001 The dosimetric consequences of intra-fraction
prostate motion in intensity modulation Med. Phys. 28 1253
Lattanzi J 2000 Application of ultrasound imaging for conformal therapy in early prostate
carcinoma Radiother. Oncol. 56 (Suppl. 1) S7
Lattanzi J, McNeeley S, Donnelly S, Palacio E, Hanlon A, Schultheiss T E and Hanks G
E 2000 Ultrasound-based stereotactic guidance in prostate cancerquantification of
organ motion and set-up errors in external beam radiation therapy Computer Aided
Surgery 5 28995
Lattanzi J, McNeeley S, Hanlon A, Schultheiss T E and Hanks G E 1999a Ultrasoundbased stereotactic guidance of precision conformal external beam radiation therapy
in clinically localized prostate cancer Urology 55 738
Lattanzi J, McNeeley S, Pinover W, Horwitz E, Das I, Schultheiss T E and Hanks G E
1999b A comparison of daily CT localization to a daily ultrasound-based system in
prostate cancer Int. J. Radiat. Oncol. Biol. Phys. 43 71925
Laub W, Alber M, Birkner M and Nusslin F 2000a Monte Carlo dose computation for
IMRT optimization Phys. Med. Biol. 45 174154
Laub W and Bakai A 2001 Monte Carlo calculations and measurements for verification of
an IMRT dose distribution in a thorax phantom Med. Phys. 28 1283
Laub W U, Bakai A and Nusslin F 2001 Intensity modulated irradiation of a thorax
phantom: comparisons between measurement, Monte Carlo calculations and pencil
beam calculations Phys. Med. Biol. 46 1695706
Laub W U, Yan D, Alber M, Nusslin F, Martinez A and Wong J 2000b IMRT in the
treatment of colo-rectal cancer: The influence of profile smoothing on the efficiency
of delivery Proc. 42nd Annual ASTRO Meeting. Int. J. Radiat. Oncol. Biol. Phys. 48
(Suppl.) 340
Lauterbach M, Siebers J, Mohan R and Wu Q 2001 IMRT optimization based on
deliverable intensities Med. Phys. 28 1204
Leal A, Sanchez-Doblado F, Capote R, Carrasco E, Lagares J I, Arrans R, Rosello J V
and Perucha M 2002 Monte Carlo based study of the influence of double-focusing on
dose distribution in IMRT treatments Radiother. Oncol. 64 (Suppl. 1) S217
Lee C, Earl M, Shepard D and Yu C 2002b Acceleration of IMRT optimisation using
objective function random sampling without replacement Med. Phys. 29 1255
Lee K Y, Chau M C and Cheung K Y 2001b Design of an inexpensive phantom for IMRT
certification Radiother. Oncol. 61 (Suppl. 1) S110
2001c Virtual pre-treatment verification for IMRT treatments Radiother. Oncol. 61
(Suppl. 1) S25
References
419
Lee M C, Xia P, Lam S C P and Ma C-M 2001a Monte Carlo dose calculations for serial
tomography Med. Phys. 28 1267
Lee N, Akazawa C, Akazawa P, Quivey J M, Tang C, Verhey L J and Xia P 2004 A forwardplanned treatment technique using multisegments in the treatment of head-and-neck
cancer Int. J. Radiat. Oncol. Biol. Phys. 59 58494
Lee N and Xia P 2003 In response to Drs Clark et al Int. J. Radiat. Oncol. Biol. Phys. 55
11523
Lee N, Xia P, Quivey J M, Sultanem K, Poon I, Akazawa C, Akazawa P, Weinberg V and
Fu K K 2002a Intensity-modulated radiotherapy in the treatment of nasopharyngeal
carcinoma: an update of the UCSF experience Int. J. Radiat. Oncol. Biol. Phys. 53
1222
Lee Y K, Bollet M, Charles-Edwards G, Flower M A, Leach M O, McNair H, Moore E,
Rowbottom C and Webb S 2003 Radiotherapy treatment planning of prostate cancer
using magnetic resonance imaging alone Radiother. Oncol. 66 20316
Lehmann J and Pawlicki T 2003 IMRT: Better plans with tuning structures Med. Phys. 30
14856
Lehmann J and Xing L 2000 Role of the number of intensity levels in IMRT Proc. World
MO-CXH-20
Lepage M, Whittaker A K, Rintoul L, Back S A J and Baldock C 2001 The relationship
between radiation-induced chemical processes and transverse relaxation times in
polymer gel dosimeters Phys. Med. Biol. 46 106174
Letourneau D, Gulam M, Yan D, Oldham M and Wong J W 2004 Evaluation of a 2D diode
array for IMRT quality assurance Radiother. Oncol. 70 199206
Letourneau D, Jaffray D, Oldham M, Sharpe M and Wong J 2002 Clinical implementation
of a kilovoltage imager for setup error verification in radiotherapy Med. Phys. 29
1243
Levendag P, Wijers O, van Sornsen de Koste J, van der Est H and Nowak P 2000 Intensitymodulated radition therapy: preliminary multi-institutional experience in head and
neck cancer Radiother. Oncol. 56 (Suppl. 1) S105
Levin S 2001 Accuray: tightly targeting tumours Windhovers In Vivo The Business and
Medicine Report 19 112
Levine R Y and Braunstein M 2002 Beam configurations for 3D tomographic intensity
modulated radiation therapy Phys. Med. Biol. 47 76587
Levitt S H and Khan F M 2002 In regard to Purdy JA, Michalski JM. Does the evidence
support the enthusiasm over 3D conformal radiation therapy and dose escalation in
the treatment of prostate cancer? (Int. J. Radiat. Oncol. Biol. Phys. 2001 51 86770)
Lewellen T 2001 PET in radiation therapy planning Med. Phys. 28 (6) 1226
Lewis D G, Spezi E and Smith C W 2001 Portal verification and dosimetry of multileaf
collimator fields by full Monte Carlo simulation Radiother. Oncol. 61 (Suppl. 1) S107
Levitt S H and Khan F M 2001 The rush to judgement: does the evidence support the
enthusiasm over three-dimensional conformal radiation therapy and dose escalation
in the treatment of prostate cancer? Int. J. Radiat. Oncol. Biol. Phys. 51 8719
Leybovich L, Dogan N, Sethi A, Krasin M and Emami B 2000a A method of correcting
dose distributions produced by IMRT planning systems Proc. World Congress of
Medical Physics and the AAPM Annual Congress, August 2000 paper MO-EBR-08
420
References
Leybovich L B, Dogan N, Sethi A, Krasin M J and Emami B 2000b Improvement

of tomographic intensity modulated radiotherapy dose distributions using periodic
shifting of arc abutment regions Med. Phys. 27 161016
Leybovich L, Sethi A, Dogan N and Glasgow G 2002 Is there any benefit in using small
ion chambers for verification of target dose delivered by IMRT? Radiother. Oncol. 64
(Suppl. 1) S44
Li J S, Boyer A L and Ma C-M 2001a Verification of IMRT dose distributions using a
water beam imaging system Med. Phys. 28 246674
Li J, Boyer A and Xing L 2000b Intensity modulated arc therapy and its relation to
fixed gantry IMRT Proc. World Congress of Medical Physics and the AAPM Annual
Li J G, Williams S S, Goffinet D R, Boyer A L and Xing L 2000a Breast-conserving
radiation therapy using combined electron and intensity-modulated radiotherapy
technique Radiother. Oncol. 55 6571
Li J and Xing L 2000b Inverse planning incorporating organ motion Proc. World Congress
of Medical Physics and the AAPM Annual Congress, August 2000 paper FR-EABR10
2000a Inverse planning incorporating organ motion Med. Phys. 27 15738
Li J S, Freedman G M, Price R, Wang L, Anderson P, Chen L, Xiong W, Yang J, Pollack A
and Ma C-M 2004c Clinical implementation of intensity-modulated tangential beam
irradiation for breast cancer Med. Phys. 31 102331
Li J S, Ozhasoglu C, Deng J, Boyer A L and Ma C-M 2000c Verification of IMRT dose
distributions using a water beam imaging system Proc. World Congress of Medical
Physics and the AAPM Annual Congress, August 2000 paper TU-EBR-01
Li J S, Yang J, Price R, Chen L and Ma C-M 2003 Machine output linearity for breast
Li J and Yu C 2001b Simultaneous optimization of beam intensity and beam orientation
for IMRT Int. J. Radiat. Oncol. Biol. Phys. 51 (Suppl. 1) 734
Li K, Dai J and Ma L 2004a Simultaneous minimizing monitor units and number of
segments without leaf end abutment for segmental intensity modulated radiation
therapy Med. Phys. 31 50712
Li X A, Ma L, Naqvi S, Shih R and Yu C 2001c Monte Carlo dose verification for intensitymodulated arc therapy Phys. Med. Biol. 46 226982
Li X A, Ma L, Yu C X, Naqvi S and Holmes T 2000d Monte Carlo dose verification for
intensity modulated arc therapy Proc. 42nd Annual ASTRO Meeting. Int. J. Radiat.
Li Y, Yao J and Yao D 2002 Direct optimization of segments for intensity-modulated
radiotherapy Med. Phys. 29 (6) 125960
2003a Aperture based optimization for static IMRT using genetic algorithm Med.
Phys. 30 1372
2003b Genetic algorithm based automatic beam angles selection in IMRT planning
Med. Phys. 30 1373
2003c Genetic algorithm based deliverable segments optimization for static intensitymodulated radiotherapy Phys. Med. Biol. 48 335374
2004b Automatic beam angle selection in IMRT planning using genetic algorithm
Lian J, Cotrutz C and Xing L 2002 IMRT dose optimization with probabilistic dose
prescription Med. Phys. 29 1260
References
421
Lian J and Xing L 2004 Biological model based IMRT optimisation with inclusion of
parameter uncertainty Proc. 14th Int. Conf. on the Use of Computers in Radiation
Lillicrap S C, Morgan H M and Shakeshaft J T 2000 X-ray leakage during radiotherapy
Brit. J. Radiol. 73 7934
Ling C 2001a Potential of biological images for radiation therapy of cancer Med. Phys. 28
1226
2001b Potential of biological images for radiation therapy of cancer Med. Phys. 28
1972
2001c Current status of intensity-modulated radiotherapy:clinical aspects Radiother.
2003 UKRO Lecture: Intensity modulated radiotherapy and biological images
Ling C C, Burman C, Chui C S, Kutcher G J, Leibel S A, LoSasso T, Mohan R, Bortfeld
T, Reinstein L, Spirou S, Wang X H, Wu Q, Zelefsky M and Fuks Z 1996 Conformal
radiation treatment of prostate cancer using inversely-planned intensity-modulated
photon beams produced with dynamic multileaf collimation. Int. J. Radiat. Oncol.
Linthout N, Verellen D, van Acker S, van de Vondel I, Coppens L and Storme G 2002
Assessment of the acceptability of the Elekta multileaf collimator (MLC) within the
CORVUS planning system for static and dynamic delivery of intensity modulated
beams (IMBs) Radiother. Oncol. 63 1214
Little D J, Dong L, Levy L B, Chandra A and Kuban D A 2003 Use of portal images and
BAT ultrasonography to measure setup error and organ motion for prostate IMRT:
Implications for treatment margins Int. J. Radiat. Oncol. Biol. Phys. 56 121824
Litzenberg D W, Moran J M and Fraass B A 2002a Incorporation of realistic delivery
limitations into dynamic MLC treatment delivery Med.Phys. 29 81020
2002b Incorporation of realistic delivery limitations into dynamic MLC treatment
delivery Med. Phys. 29 1267
Liu C and Xing L 2000 Dosimetric effects of mechanical inaccuracy of MLC leaf positions
on IMRT Proc. World Congress of Medical Physics and the AAPM Annual Congress,
August 2000 paper MO-GBR-01
Liu H H, Verhaegen F and Dong L 2001a Monte Carlo simulation of dynamic multi-leaf
collimators Radiother. Oncol. 61 (Suppl. 1) S40
2001b A method of simulating dynamic multileaf collimators using Monte Carlo
techniques for intensity modulated radiation therapy Phys. Med. Biol. 46 228398
Liu H H, Wang X, Dong L, Wu Q, Liao Z, Stevens C W, Guerrero T M, Komaki R, Cox
J D and Mohan R 2004 Feasibility of sparing lung and other thoracic structures with
intensity-modulated radiotherapy for non-small-cell lung cancer Int. J. Radiat. Oncol.
Biol. Phys. 58 126879
Liu T, Lizzi F L, Feleppa E J and Kutcher G J 2002 Ultrasonic tissue typing using spectrum
analysis technique and the potential role in conformal radiation therapy Radiother.
Liu W-C, Schulder M, Narra V, Kalnin A J, Cathcard C, Jacobs A, Lange G and Holodny
A I 2000 Functional magnetic resonance imaging aided radiation treatment planning
Med. Phys. 27 156372
Livsey J, Mott J H, Wylie J P, Cowan R A, Khoo V S and Logue J P 2003 IMRT for
carcinoma of the prostate Clinical Oncol. 15 (Suppl. 2) S6
422
References
Llacer J, Deasy J O, Bortfeld T R, Solberg T D and Promberger C 2003 Absence of

multiple local minima effects in intensity modulated optimization with dose-volume
constraints Phys. Med. Biol. 48 183210
Llacer J, Solberg T D and Promberger C 2001 Comparative behaviour of the dynamically
penalized likelihood algorithm in inverse radiation therapy planning Phys. Med. Biol.
46 263763
Locke J, Low D A, Grigireit T and Chao K S C 2002 Potential of tomotherapy for total
scalp treatment Int. J. Radiat. Oncol. Biol. Phys. 52 5539
Lofroth P-O, Bergstrom P, Brannstrom K, Nystrom E, Zackrisson B and Widmark A 2000
Conformal radiotherapy of the prostate with high precision Radiother. Oncol. 56
(Suppl. 1) S108
Logue J 2000 Clinical implementation of IMRT Radiother. Oncol. 56 (Suppl. 1) S86
Loi G, Manfredda I, Secco C, Sacchetti G, Mones E, Natrella M, Bagnasacco P, Inglese E,
Cotroneo A and Krengli M 2002 Fusion of CT-MR-SPECT images for the delineation
of target volume in the treatment plan of high-grade gliomas. Preliminary results
Loof M, Engstrom P, Uttman K and Nystrom H 2001 Optimisation of IMRT: can
geometrical uncertainties be incorporated into the optimisation process? Radiother.
Lomax A J 2001 Methods for assessing the effects of inter- and intra-fraction positional
errors in radiotherapy Radiother. Oncol. 61 (Suppl. 1) S22
Lopez Torrecilla J 2001 The introduction of IMRT in the clinic Radiother. Oncol. 61
(Suppl. 1) S31
LoSasso T L 2003 IMRT delivery system QA IMRTthe State of the Art: AAPM Medical
LoSasso T, Burman C, Chui C and Ling C 1998a QA and verification of intensity
modulation radiotherapy. Med. Phys. 25 A206
LoSasso T, Chui C-S, and Ling C C 1998b Physical and dosimetric aspects of a multileaf
collimation system used in the dynamic mode for implementing intensity modulated
radiotherapy. Med. Phys. 25 191927
2001 Comprehensive quality assurance for the delivery of intensity modulated
radiotherapy with a multileaf collimator used in the dynamic mode Med. Phys. 28
220919
Lotz H, Remeijer P, van Herk M, Lebesque J, de Bois J and Zijp L 2003 A model to predict
the bladder shape from changes in bladder and rectal volume Radiother. Oncol. 68
(Suppl. 1) S35
Love P A, Evans P M, Leach M O and Webb S 2003 Polymer gel measurement of
dose homogeneity in the breast comparing MLC intensity modulation with standard
wedged dosimetry Phys. Med. Biol. 48 106574
Low D A 2001 Verification methods for IMRT Med. Phys. 28 1271
2003 Radiation shielding for IMRT IMRTthe State of the Art: AAPM Medical
Low D A, Bradley J D, Deasy J O, Laforest R, Parikh P J, Sohn J W, Hibbard L S,
Dehdashti F and Mutic S 2001b Lung trajectory mapping Int. J. Radiat. Oncol. Biol.
Phys. 51 (Suppl. 1) 2089
References
423
Low D A, Dempsey J F, Markman J, Mutic S, Klein E E, Sohn J W and Purdy J A 2002a

Toward automated quality assurance for intensity-modulated radiation therapy Int. J.
Low D A, Dempsey J F, Markman J, Mutic S, Williamson J F and Purdy J A 2000b
Applicator-guided intensity modulated radiation therapy Proc. 42nd Annual ASTRO
Low D A, Grigsby P W, Dempsey J F, Mutic S, Williamson J F, Markman J, Chao K S
C, Klein E E and Purdy J A 2002b Applicator guided intensity modulated radiation
therapy Int. J. Radiat. Oncol. Biol. Phys. 52 14006
Low D A, Markman J, Dempsey J F and Mutic S 2000a Noise in polymer gel measurements
using MRI Med. Phys. 27 181417
Low D A, Parikh P J, Nystrom M, El Naqa I M, Nystrom M M, Lu W, Hubenschmidt J P,
Wahab S H, Mutic S, Singh A K, Christensen G and Bradley J D, 2004 Quantitative
4-D CT using a multislice CT scanner Proc. 14th Int. Conf. on the Use of Computers
in Radiation Therapy (Seoul, May) pp 5761
Low D A, Sohn J W, Klein E E, Markman J, Mutic S and Dempsey J F 2001a
Characterization of a commercial multileaf collimator used for intensity modulated
Lu T-X, Mai W-Y, Teh B S, Zhao C, Han F, Huang Y, Deng X-W, Lu L-X, Huang S-M,
Zeng Z-F, Lin C-G, Lu H H, Chiu J K, Carpenter L S, Grant W H III, Woo S Y, Cui
N-J and Butler E B 2004 Initial experience using intensity-modulated radiotherapy
for recurrent nasopharyngeal carcinoma Int. J. Radiat. Oncol. Biol. Phys. 58 6827
Lu W and Mackie T R 2002 Tomographic motion detection and correction directly in
sinogram space Phys. Med. Biol. 47 126784
Lu X-Q, Burman C, Mychaiczak B, Hirsch A, Chui C-S, Leibel S and Ling C 2000
Feasibility study of using low-energy intensity modulated radiation for the treatment
of localized prostate carcinoma to high doses Proc. World Congress of Medical
Physics and the AAPM Annual Congress, August 2000 paper MO-EBR-03
Lu Y, Spelbring R and Chen G T Y 1997 Functional dose-volume histograms for
functionally heterogeneous normal organs Phys. Med. Biol. 42 34556
Luan S, Wang C, Chen D Z, Hu X S, Naqvi S A, Yu C X and Lee C L 2004 A new MLC
leaf segmentation algorithm/software for step-and-shoot IMRT delivery Med. Phys.
31 695707
Luan S, Wang C, Naqvi S A, Chen D Z, Hu X S, Lee C L and Yu C X 2003 A new leaf
sequencing algorithm/software for step and shoot IMRT delivery Med. Phys. 30 1404
Lujan A E, Balter J M and Ten Haken R K 2003 A method for incorporating organ motion
due to breathing into 3D dose calculations in the liver: Sensitivity to variations in
motion Med. Phys. 30 26439
Lujan A E, Roeske J C, Johnson L S, Reft C S, Laughton C, Kung J H and Pelizzari C
A 2001a Verification of IMRT dose calculations using an independent monitor unit
calculation Med. Phys. 28 1282
Lujan A E, Roeske J C and Mundt A J 2001b Intensity-modulated radiation therapy as
a means of reducing dose to bone marrow in gynaecologic patients receiving whole
pelvic radiation therapy Int. J. Radiat. Oncol. Biol. Phys. 51 (Suppl. 1) 220
Luo C, Wu X, Chen Z, Shao H, Han H, Wolfson A and Markoe A 2000 Reducing dose
to the contralateral breast in the conventional two-wedge tangential breast or chest
wall radiotherapy Proc. World Congress of Medical Physics and the AAPM Annual
Congress, August 2000 paper MO-FXH-52
424
References
Luo C, Wu X, Shao H and Markoe A 2002 Linear algebraic approach for IMRT inverse
planning Med. Phys. 29 1260
Luxton G, Hancock S L and Boyer A L 2004 Dosimetry and radiobiologic model
comparison of IMRT and 3D conformal radiotherapy in treatment of carcinoma of
the prostate Int. J. Radiat. Oncol. Biol. Phys. 59 26784
Luxton G, Hancock S L, Chen Y, Xing L and Boyer A L 2000 Reduction of bowel dose in
lymph node irradiation with IMRT treatment of prostate cancer Proc. 42nd Annual
Lyatskaya Y, Chin L, Harris J and Lu H 2002 Accurate patient position control for breast
cancer treatment using respiratory manoeuvres Med. Phys. 29 1239
Ma C 2000 Monte Carlo for treatment planning and IMRT dose verification Radiother.
2003 Why Cyberknife? Radiother. Oncol. 68 (Suppl. 1) S39
Ma C-M, Ding M, Li J S, Lee M C, Pawlicki T and Deng J 2003c A comparative dosimetric
study on tangential photon beams, intensity-modulated radiation therapy (IMRT) and
modulated electron radiotherapy (MERT) for breast cancer treatment Phys. Med.
Biol. 48 90024
Ma C M, Guerrero T, Pawlicki T, Luxton G and Adler J 2001b Treatment planning and
dosimetry verification for the Cyberknife system Radiother. Oncol. 61 (Suppl. 1) S16
Ma C-M, Jiang S B, Pawlicki T, Chen Y, Li J S, Deng J and Boyer A L 2003a A quality
assurance phantom for IMRT dose verification Phys. Med. Biol. 48 56172
Ma C, Jiang T, Pawlicki T, Chen Y, Li J, Deng J, Lee M, Mok E and Boyer A 2000a
A quality assurance phantom for IMRT dose verification Proc. World Congress of
Medical Physics and the AAPM Annual Congress, August 2000 paper TU-EBR-06
Ma C, Li J, Chen L, Wang L, Price R, McNeeley S, Ding, M and Fourkal E 2002b Monte
Carlo dose verification for advanced radiotherapy treatment techniques Radiother.
Ma C M, Li J S, Pawlicki T, Jiang S B, Deng J, Lee M C, Koumrian T, Luxton M and Brain
S 2002a A Monte Carlo dose calculation tool for radiotherapy treatment planning
Ma C-M, Pawlicki T, Jiang S B, Li J S, Deng J, Mok E, Kapur A, Xing L, Ma L and Boyer
A L 2000c Monte Carlo verification of IMRT dose distributions from a commercial
treatment planning optimization system Phys. Med. Biol. 45 248395
Ma C-M, Pawlicki T, Lee M C, Jiang S B, Li J S, Deng J, Yi B, Mok E and Boyer A L
2000e Energy and intensity-modulated electron beams for radiotherapy Phys. Med.
Biol. 45 2293311
Ma C M, Pawlicki T, Li J S, Deng J, Shahine B, Lee M C and Boyer A L 2001a Quality
assurance phantoms and Monte Carlo dose verification Radiother. Oncol. 61 (Suppl.
1) S4
Ma L 2002a Smoothing intensity-modulated radiotherapy delivery under hardware
constraints Med. Phys. 29 1304
2002b Smoothing intensity-modulated treatment delivery under hardware constraints,
Med. Phys. 29 293745
Ma L, Phaisangittisakul N, Yu C X and Sarfaraz M 2003b A quality assurance method for
analysing and verifying intensity modulated fields Med. Phys. 30 20828
Ma L, Yu C and Sarfaraz M 2000b A dosimetric leaf-setting strategy for shaping radiation
fields using a multileaf collimator Med. Phys. 27 9727
References
425
Ma L, Yu C, Sarfaraz M, Holmes T, Shepard D, Li A, DiBiase S, Amin P, Suntharalingam

M and Mansfield C 2000d Simplified intensity modulated arc therapy for prostate
cancer treatments Proc. 42nd Annual ASTRO Meeting. Int. J. Radiat. Oncol. Biol.
Phys. 48 (Suppl.) 350
MacDougall N D, Pitchford W G and Smith M A 2002 A systematic review of the precision
and accuracy of dose measurements in photon radiotherapy using polymer and Fricke
MRI gel dosimetry Phys. Med. Biol. 47 R10721
2003 Letter to the Editor: Reply to Comment on A systematic review of the precision
and accuracy of dose measurements in photon radiotherapy using polymer and Fricke
MRI gel dosimetry Phys. Med. Biol. 48 L1922
MacKay R I and Amer A 2000 Radiobiological analysis of 3D motion Radiother. Oncol.
56 (Suppl. 1) S51
MacKay R I, Graham P A, Logue J P and Moore C J 2000 Patient positioning using detailed
3D surface data for patients undergoing conformal radiotherapy for carcinoma of the
prostate: a feasibility study Int. J. Radiat. Oncol. Biol. Phys. 49 22530
MacKenzie M A and Robinson D M 2002 Intensity modulated arc deliveries approximated
by a large number of fixed gantry position sliding window dynamic multileaf
collimator fields Med. Phys. 29 235965
Mackie T 2001 Tomotherapy Med. Phys. 28 1299
Mackie T R, Balog J, Becker S, Tome W, Reckwerdt P J, Kapatoes J, Ruchala K, Smilowitz
J and Olivera G 2002 Can helical tomotherapy do simple radiotherapy procedures?
Med. Phys. 29 1211
Mackie T R, Kapatoes J, Ruchala K, Lu W, Wu C, Olivera G, Forrest L, Tome W, Welsh J,
Jeraj R, Herari P, Reckwerdt P, Paliwal B, Ritter M, Keller H, Fowler J and Mehta M
2003a Image guidance for precise conformal radiotherapy Int. J. Radiat. Oncol. Biol.
Phys. 56 89105
Mackie T R, Olivera G H, Kapatoes J M, Ruchala K J, Balog J R, Tome W A, Hui S,
Kissick M, Wu C, Jeraj R, Reckwerdt P J, Harari P, Ritter M, Forrest L, Welsh J and
Mehta M P 2003b Helical tomotherapy IMRTthe State of the Art: AAPM Medical
Mackie T R, Ruchala K, Olivera G, Kapatoes J, Lu W, Fang G, Murray D, Reckwerdt P,
Hughes J, Kissick M and Jeraj R 2004 Overview of helical tomotherapy Proc. 14th
Macklis R, Weinhous M and Harnisch G 2000 Intensity-modulated radiotherapy:
rethinking basic treatment planning paradigms Int. J. Radiat. Oncol. Biol. Phys. 48
31718
Macpherson M, Crljenko T, Vo D, El-Hakim S and Gerig L 2002 A machine-vision based
radiotherapy gating device Med. Phys. 29 1213
Maes A, Weltens C, Huyskens D and van den Bogaert W 2002 Conformal and intensitymodulated radiotherapy for preservation of salivary function Radiother. Oncol. 64
(Suppl. 1) S110
Maes A, Weltens C, van Esch A, Kutcher G, Huyskens D and van den Bogaert W 2000
The potential of IMRT for parotid sparing in head and neck cancer Radiother. Oncol.
56 (Suppl. 1) S109
Mageras G S 2001 Respiratory gating in conformal radiotherapy Radiother. Oncol. 61
(Suppl. 1) S35
426
References
2002 Respiration correlated CT techniques for gated treatment of lung cancer

2004 When is a deep inspiration breath-hold technique during external beam
radiotherapy beneficial? Proc. 14th Int. Conf. on the Use of Computers in Radiation
Therapy (Seoul, May) p 52
Mageras G, Joshi S, Davis B, Pevsner A, Hertanto A, Yorke E, Rosenzweig K, Erdi Y,
Nehmeh S, Humm J, Larson S M and Ling C C 2004b Evaluation of an automated
deformable matching method for quantifying lung tumor motion in respiration
correlated CT images Proc. 14th Int. Conf. on the Use of Computers in Radiation
Mageras G S, Sillanpaa J, Seppi E, Chang J, Ling C C and Amols H 2004a Preclinical
studies of megavoltage cone beam CT systems for tumor localization in gated
radiotherapy of lung cancer Proc. 14th Int. Conf. on the Use of Computers in
Mageras G S and Yorke E 2004 Deep inspiration breath hold and respiratory gating
strategies for reducing organ motion in radiation treatment Seminars Radiat. Oncol.
14 6575
Mah D, Freedman G, Milestone B, Movsas B, Mitra R, Horwitz E and Hanks G E 2001
Dancing prostates: Measurement of intra-fractional prostate motion using MRI Int.
J. Radiat. Oncol. Biol. Phys. 51 (Suppl. 1) 212
Mah D, Hanley J, Rosenzweig K E, Yorke E, Braban L, Ling C C, Leibel S A and
Mageras G 2000 Technical aspects of the deep inspiration breath-hold technique in
the treatment of thoracic cancer Int. J. Radiat. Oncol. Biol. Phys. 48 117585
Mah K, Danjoux C E and Caldwell C B 2000 Is spiral CT too fast for radiation therapy
planning of thoracic neoplasms? Radiother. Oncol. 56 (Suppl. 1) S38
Malden C H, Meehan C A, Mubata C D and Bidmead A M 2001 Dose verification of
dynamic IMRT for treatment of the prostate and pelvic nodes Radiother. Oncol. 61
(Suppl. 1) S108
Malsch U, Fruhling C and Bendl R 2004 Methods for elastic adaptation of segmented
volumes of interest for adaptive radiotherapy planning Proc. 14th Int. Conf. on the
Use of Computers in Radiation Therapy (Seoul, May) pp 6973
Manning M A, Wu Q, Cardinale R M, Mohan R, Lauve A D, Kavanagh B D, Morris M
M and Schmidt-Ullrich R K 2001 The effect of setup uncertainty on normal tissue
sparing with IMRT for head-and-neck cancer Int. J. Radiat. Oncol. Biol. Phys. 51
14009
Manning M A, Wu Q, Mohan R, Cardinale R M, Kavanagh B D, Morris M M and SchmidtUllrich R K 2000 The effect of set-up uncertainty on normal tissue sparing with IMRT
for head and neck Proc. 42nd Annual ASTRO Meeting. Int. J. Radiat. Oncol. Biol.
Phys. 48 (Suppl.) 193
Mapelli M, Fumagalli M, Pellegrini R, Salehi N, Milanesi I and Fariselli L 2001
Physical and dosimetric characterization of a micro-multileaf collimator for dynamic
conformal radiotherapy Radiother. Oncol. 61 (Suppl. 1) S97
Marcie S, Martin E, Dalmasso C, Bensadoun R J, Serrano B, Herault J, Costa A,
Bensadoun R J, Hachem S and Gerard J P 2002 IMRT implementation at the NICECAL: from theoretical to clinical cases Radiother. Oncol. 64 (Suppl. 1) S215
Markman J, Low D A, Beavis A W and Deasy J O 2002 Beyond bixels: Generalizing
the optimisation parameters for intensity modulated radiation therapy. Med. Phys. 29
2298304
References
427
Martens C, Budgell G J and Claus F 2001a Improved delivery efficiency for step and shoot
intensity-modulated radiotherapy using a fast-tuning magnetron Radiother. Oncol. 61
(Suppl. 1) S83
Martens C, Claeys I, de Wagter C and de Neve 2002 The value of radiographic film for the
characterization of intensity-modulated beams Phys. Med. Biol. 47 222134
Martens C, de Gersem W, de Neve W and de Wagter C 2001b Combining the advantages
of step-and-shoot and dynamic delivery of intensity-modulated radiotherapy by
interrupted dynamic sequences Int. J. Radiat. Oncol. Biol. Phys. 50 54150
Martens C, Reynaert N, de Wagter C, Nilsson P, Palmans H, Thierens H and de Neve W
2001c Underdosage of the mucosa for small fields as used in IMRT: a comparison
between radiochromic film measurements, Monte Carlo simulations, and collapsed
cone convolution calculations Radiother. Oncol. 61 (Suppl. 1) S25
Martin E and Hachem A 2003 Validation of intensity modulation on a commercial
treatment planning system Med. Phys. 30 92536
Martineau A and Manens J P 2001 Dosimetry of small photon fields for conformal
radiotherapy Phys. Medica. 17 113
Martinez A A, Yan D, Lockman D, Brabbins D, Kota K, Sharpe M, Jaffray D A, Vicini
F and Wong J 2001 Improvement in dose escalation using the process of adaptive
radiotherapy combined with three-dimensional conformal or intensity-modulated
beams for prostate cancer Int. J. Radiat. Oncol. Biol. Phys. 50 122634
Mather M L, Whittaker A K and Baldock C 2002 Ultrasound evaluation of polymer gel
dosimeters Phys. Med. Biol. 47 144958
Mavroidis P, Lind B K, van Dijk J, Koedooder K, de Neve W, de Wagter C, Planskoy
B, Rosenwald J-C, Proimos B, Kappas C, Danciu C, Benassi M, Chierego G and
Brahme A 2000 Comparison of conformal radiation therapy techniques within the
dynamic radiotherapy project Dynarad Phys. Med. Biol. 45 245981
Mayles W P M, Wolff T, Cassapi L, Clements R W, Fenwick J D, Gately A, Martin L,
and Mayles H 2002 Introduction of IMRT as a routine clinical treatment Proc. IPEM
Biennial Meeting (Southampton, 1617 July) pp 1112
Mazurier J, Castelain B and Lartigau E 2001a Dosimetric evaluation of IMRT at Centre
Oscar Lambret Radiother. Oncol. 61 (Suppl. 1) S64
2001b Dosimetric evaluation of intensity modulated treatments in radiotherapy at
Oscar Lambret Centre Phys. Medica. 17 114
Mazurier J, Kouto H, Poupon L, Castelain B and Lartigau E 2002 Implementation and
quality control of IMRT for head and neck cancer Radiother. Oncol. 64 (Suppl. 1)
S215
McBain C A, Sykes J S, Buckley D L, Amer A, Moore C J, Cowan R A and Khoo V
S 2003 Optimizing bladder radiotherapy: MR assessment of time-dependent organ
motion Clinical Oncol. 15 (Suppl. 2) S13
McClean B and Rock L 2001 Characterisation of small MLC-defined fields for use in
IMRT treatment techniques Phys. Medica. 17 161
McCurdy B, Malkoske K and Furutani K 2002 An automated compensator exchanging
(ACE) device to facilitate the cost-effective delivery of IMRT Med. Phys. 29 1306
McKenzie A L 2000 How should breathing motion be combined with other errors when
drawing margins around clinical target volumes? Brit. J. Radiol. 73 9737
2003a Optimum radiotherapy planning margins RAD Magazine 29 (334) 312
2003b Protocols for target volumes and organs at risk Clinical Oncol. 15 (Suppl. 2)
S12
428
References
McKenzie A L, van Herk M and Mijnheer B 2000 The width of margins in radiotherapy
plans Phys. Med. Biol. 45 333142
2002 Margins for geometric uncertainty around organs at risk in radiotherapy
McNair H 2002 Clinical IMRT: Impact on a radiotherapy department Radiother. Oncol. 64
(Suppl. 1) S79
McNair H A, Adams E J, Clark C H, Miles E A and Nutting C M 2003b Implementation
of IMRT in the radiotherapy department Brit. J. Radiol. 76 8506
McNair H, Christian J, Knowles C, Symonds-Tayler J and Brada M 2003a Can the
active breathing co-ordinator be used for patients receiving radical lung radiotherapy?
McNair H A, Francis G and Balyckyi J 2004 Clinical implementation of dynamic intensitymodulated radiotherapy: radiographers perspectives Brit. J. Radiol. 77 4938
McNeeley S, Buyyounouski M, Price R, Horwitz E and Ma C 2003 The use of ultrasound
in target localization Radiother. Oncol. 68 (Suppl. 1) S16
McShan D, Lynn K, Vineberg K and Fraass B 2002 Radiotherapy plan optimization
accounting for set-up and motion uncertainty using a multiple instance geometry
approximation Med. Phys. 29 1257
McShan D, Ten Haken R, Kessler M, Balter J, Lam K and Fraass B 2001 Radiotherapy plan
optimisation accounting for set-up and motion uncertainty using a multiple instance
geometry approximation Radiother. Oncol. 61 (Suppl. 1) S13
Mechalakos J G, Mageras G S, Zelefsky M J, Lyass O, van Herk M, Kooy H M, Leibel S
A and Ling C C 2002 Time trends in organ position and volume in patients receiving
prostate three-dimensional conformal radiotherapy Radiother. Oncol. 62 2615
Mechalakos J G, Yorke E, Mageras G S, Hertano A, Jackson A, Obcemea C, Rozenzweig
K and Ling C C 2004 Dosimetric effect of respiratory motion in external beam
radiotherapy of the lung Radiother. Oncol. 71 191200
Meedt G, Alber and Nuesslin F 2001 A method of 3D beam direction optimisation in IMRT
2002 Beam direction optimisation in head and neck IMRT Radiother. Oncol. 64
(Suppl. 1) S1011
2003 Non-coplanar beam direction optimisation for intensity-modulated radiotherapy
Phys. Med. Biol. 48 29993019
Meeks S L, Tome W A, Bouchet L G, Haas A C, Buatti J M and Bova F J 2003 Patient
positioning using optical and ultrasound techniques IMRT- the state of the art: AAPM
Meyer J 2003 Beam direction optimization for anatomy-based IMRT using beams-eyeview metrics Clinical Oncol. 15 (Suppl. 2) S30
Meyer J, Mills J A, Haas O C L, Burnham K J and Parvin E M 2001 Accommodation
of couch constraints for coplanar intensity modulated radiation therapy Radiother.
Oncol. 61 2332
Meyer J, Phillips M H, Cho P S, Kalet I and Doctor J N 2004 Application of influence
diagrams to prostate intensity-modulated radiation therapy plan selection Phys. Med.
Biol. 49 163753
Miften M, Das S, Su M and Marks L 2004a Functional equivalent uniform dose (fEUD)
for reporting and comparing functional image-guided dose distributions Proc. 14th
References
429
2004b Incorporation of functional imaging data in the evaluation of dose distributions

using the generalized concept of equivalent uniform dose Phys. Med. Biol. 49 1711
21
Mihailidis D and Gibbons J 2002 Optimum number of treatment fields for inversely
planned prostate irradiation: a planning comparison Med. Phys. 29 1335
Mijnheer B 2001a Possibilities and limitations of the use of radiographic film for the
verification of IMRT fields Radiother. Oncol. 61 (Suppl. 1) S3
Miles E A, Clark C H, Guerrero Urbano M T, Dearnaley D P and Nutting C M 2003 How
routine can IMRT become in daily clinical practice? Clinical Oncol. 15 (Suppl. 2)
S30
Miller T R and Grigsby P W 2003 In response to Dr. Beitler Int. J. Radiat. Oncol. Biol.
Phys. 55 282
Minken A W H and Mijnheer B J 2001 Possibilities and limitations in the use of
radiographic film for the evaluation of dose delivery in small fields Med. Phys. 28
1282
Mitra R, Bayouth J and Das I 2001a Startup characteristics and dosimetry of small monitor
unit (MU) segments in step and shoot IMRT delivery for two types of digital linear
accelerators Med. Phys. 28 1202
Mitra R, Mah D, Das I, Horwitz E and Hanks G 2001b Evaluation of dosimetric
characteristics of a virtual micro-multileaf collimatorSiemens HD270 MLC Med.
Phys. 28 1277
Mittal B, Kepka A, Mahadevan A, Kies M, Pelzer H, List M, Rademaker A and Logemann
J 2001 Use of IMRT to reduce toxicity from concomitant radiation and chemotherapy
for advanced head and neck cancers Int. J. Radiat. Oncol. Biol. Phys. 51 (Suppl. 1)
823
Mock U, Kodym R, Weitmann H, Wolff U, Jetzek S, Pittner I and Potter R 2000 Influence
of set-up displacements in conformal radiotherapy of prostate cancer on changes
in dose-volume histograms (DVH), tumour control probabilities (TCP) and normal
tissue complication probabilities (NTCP) of the bladder and rectum Radiother. Oncol.
56 (Suppl. 1) S109
Moeckli R, Jeanneret Sozzi W, Valley J F and Mirimanoff R O 2002 The reasons for
discrepancies in target volume delineation: 1. Physical part Radiother. Oncol. 64
(Suppl. 1) S2089
Mohan R 2002 IMRT present and future prospects Radiother. Oncol. 64 (Suppl. 1) S64
Mohan R, Antolak J and Hendee W R 2001a Monte Carlo techniques should replace
analytical methods for estimating dose distributions in radiotherapy treatment
planning Med. Phys. 28 1236
Mohan R, Arnfield M, Tong S, Wu Q and Siebers J 2000 The impact of fluctuations in
intensity patterns on the number of monitor units and the quality and accuracy of
intensity modulated radiotherapy Med. Phys. 27 122637
Mohan R, Wang X, Jackson A, Bortfeld T, Boyer A L, Kutcher G J, Leibel S A, Fuks Z
and Ling C C 1994 The potential and limitations of the inverse radiotherapy technique
Radiotherapy and Oncology 32 23248
Mohan R, Wu Q, Morris M, Lauve A, Tong S, Benedict S and Schmidt-Ullrich 2001b
Simultaneous integrated boost (SIB) IMRT of advanced head and neck squamous
cell carcinomasdosimetric analysis Int. J. Radiat. Oncol. Biol. Phys. 51 (Suppl. 1)
1801
430
References
Moore C J and Graham P A 2000 3D dynamic surface sensing and CT-body matching:
a tool for patient set-up and monitoring in radiotherapy Computer Aided Surgery 5
23445
Moran J M, Fraass B A, Lam K L and Litzenberg D L 2001a An integrated system for
DMLC (IMRT) fluence verification Med. Phys. 28 1284
Moran J M, Litzenberg D L, Nurushev T S and Fraass B A 2001b Dosimetric comparisons
of static, SMLC, and DMLC IMRT delivery techniques Radiother. Oncol. 61 (Suppl.
1) S17
Morgan A 2001 Commissioning and implementation of a simple IMRT technique
Abstracts of the Elekta IMRT User Meeting 2001 in NKI Amsterdam, p 14
Morita K 2000 Keynote addressconformal RT and conformation RT Int. J. Radiat.
Mornex F, Chapet O, Chapuy S, Dimcovski Z, Morignat E, Romestaing P, Sentenac I and
Gerard J P 2000 IRIS, A dynamic portal imaging system to visualize and control
treatment in radiotherapy Radiother. Oncol. 56 (Suppl. 1) S61
Morr J, DiPetrillo T, Tsai J-S, Engler M and Wazer D E 2002 Initial experience with
ultrasound localization for positioning prostate cancer patients for external beam
radiotherapy. Int. J. Radiat. Oncol. Biol. Phys. 53 11249
Morr J, DiPetrillo T, Tsai J and Wazer D 2000 Clinical use of daily transabdominal
ultrasound localization with intensity modulated radiation therapy (IMRT) for
prostate cancer Proc. 42nd Annual ASTRO Meeting. Int. J. Radiat. Oncol. Biol. Phys.
48 308
Morrill S, Bayouth J and Mehta S 2001 A comparison of inverse IMRT treatment plans
Med. Phys. 28 (Suppl.) 1206
Mosleh-Shirazi M A 2002 Measurement of the field shaping properties of the Elekta highresolution MLC (beam modulator) Radiother. Oncol. 64 (Suppl. 1) S21213
Mosleh-Shirazi M A, Evans P M, Swindell W, Webb S and Partridge M 1998 A conebeam megavoltage CT scanner for treatment verification in conformal radiotherapy.
Mott J H L, Budgell G F and Williams P C 2001 Treatment planning for IMRT using
pre-determined intensity modulations for particular target geometriesa feasibility
study Radiother. Oncol. 61 (Suppl. 1) S32
Mott J H, Livsey J E and Logue J P 2004 Development of a simultaneous boost IMRT class
solution for hypofractionated prostate cancer protocol Brit. J. Radiol. 77 37786
Moya Garcia F 2001 The need for IMRT: a PET point of view Radiother. Oncol. 61 (Suppl.
1) S1
Muench R K, Seiler P G and Verwey J 2001 Application of magnetic tracking for the
continuous precise measurement of tumour positions in conformal radiotherapy Phys.
Medica. 17 164
Mundt A J, Lujan A E, Rotmensch J, Waggoner S E, Yamada S D, Fleming G and Roeske
J C 2002 Intensity-modulated whole pelvic radiotherapy in women with gynecologic
malignancies Int. J. Radiat. Oncol. Biol. Phys. 52 13307
Mundt A J, Roeske J C and Lujan A E 2001 Clinical experience with intensitymodulated whole pelvic radiation therapy (IM-WPRT) in patients with gynaecologic
malignancies Int. J. Radiat. Oncol. Biol. Phys. 51 (Suppl. 1) 220
Munro P, Seppi E, Johnsen S, Shapiro E, Tognina C, Jones D, Pavkovich J, Webb C, Mollov
I, Partain L and Colbeth R E 2002 Megavoltage cone-beam computed tomography
using a high quantum efficiency image receptor Med. Phys. 29 1340
References
431
Munter M W, Thilmann C, Hof H, Didinger B, Rhein B, Nill S, Schlegel W, Wannenmacher

M and Debus J 2003 Stereotactic intensity modulated radiation therapy and inverse
treatment planning for tumours of the head and neck region: clinical implementation
of the step and shoot approach and first clinical results Radiother. Oncol. 66 31321
Murphy M 2002 Image-guided radiosurgery using the Cyberknife Radiother. Oncol. 64
(Suppl. 1) S22
2004 Tracking moving organs in real time Seminars Radiat. Oncol. 14 91100
Murphy M J, Chang S D, Gibbs I C, Le Q-T, Hai J, Kim D, Martin D P and Adler J R
2003 Patterns of patient movement during frameless image-guided radiosurgery Int.
Murphy M J, Isaksson M, Jalden J and Ozhasoglu C 2001 Realtime compensation of
breathing motion during radiotherapy Int. J. Radiat. Oncol. Biol. Phys. 51 (Suppl.
1) 388
Murshed H, Liu H H, Liao Z, Barker J L, Wang X, Tucker S L, Chandra A, Guerrero
T, Stevens C, Change J Y, Jeter M, Cox J D, Komaki R and Mohan R 2004 Dose
and volume reduction for normal lung using intensity-modulated radiotherapy for
advanced-stage non-small-cell lung cancer Int. J. Radiat. Oncol. Biol. Phys. 58 1258
67
Mutic S, Low D A, Klein E E, Dempsey J F and Purdy J A 2001 Room shielding for
intensity-modulated radiation therapy treatment facilities Int. J. Radiat. Oncol. Biol.
Phys. 50 23946
Mutic S, Malyapa R S, Grigsby P W, Dehdashti F, Miller T R, Zoberi I, Bosch W R,
Esthappan J and Low D A 2003 PET-guided IMRT for cervical carcinoma with
positive para-aortic lymph nodesa dose-escalation treatment planning study Int.
Myrianthopoulos L, Turian J, Hamilton R and Baird C 2001 Dose rate effects on the
calculated fluence maps of IMRT fields Med. Phys. 28 1203
Nabid A, Bujold R, Benard F, Verreault J and Turgeon P P 2002 Evaluation of the size of
tumours reported by total body PET scans and pathology reports for resected tumors
Nahum A E. 2003 Tumour local-control probability prediction (TLCP) and other
radiobiological issues in the use of IMRT Radiother. Oncol. 68 (Suppl. 1) S5
Nakagawa K, Aoki Y, Akanuma A, Onogi Y, Terahara A, Sakata K, Muta N, Sasaki
Y, Kawakami H and Hanakawa K 1994 Real-time beam monitoring in dynamic
conformation therapy Int. J. Radiat. Oncol. Biol. Phys. 30 12338
Nakagawa K, Aoki Y, Tago M, Terahara A and Ohtomo K 2000a Megavoltage CT-assisted
stereotactic radiosurgery for thoracic tumours: original research in the treatment of
thoracic neoplasms Int. J. Radiat. Oncol. Biol. Phys. 48 44957
Nakagawa K, Aoki Y and Ohtomo K 2000b Dynamic conical conformal therapy using a Carm mounted accelerator Proc. 42nd Annual ASTRO Meeting. Int. J. Radiat. Oncol.
Naqvi S A, Holmes T W, Sarfaraz M, Li X A and Yu C X 2000 Improving the accuracy
of monitor unit calculation for complex leaf sequences in IMRT delivery Proc. 42nd
Annual ASTRO Meeting. Int. J. Radiat. Oncol. Biol. Phys. 48 (Suppl.) 221
Naqvi S, Holmes T and Yu C 2001a Verifying static-to-dynamic conversion of IMRT leaf
sequences by raytracing Med. Phys. 28 1268
432
References
Naqvi S A, Sarfaraz M, Holmes T, Yu C X and Li X A 2001b Analysing collimator

structure effects in head-scatter calculations for IMRT class fields using scatter
raytracing Phys. Med. Biol. 46 200928
Nederveen A J, van der Heide U A, Dehnad H, Jeroen R, van Moorselaar A, Hofman P and
Lagendijk J J W 2002 Measurements and clinical consequences of prostate motion
during a radiotherapy fraction Int. J. Radiat. Oncol. Biol. Phys. 53 20614
Neicu T, Papiez L and Jiang S B 2003 Synchronized moving aperture radiation therapy
(SMART): Dynamic MLC leaf sequencing with tumor motion Med. Phys. 30 1405
Neicu T, Shirato H, Seppenwold Y and Jiang S B 2002 Gated motion adaptive therapy
(GMAT): average tumour trajectory for lung patients Med. Phys. 29 130910
Nguyen N, Higgins P and Alaei P 2002 Evaluation of Kodak EDR2 film and its application
in IMRT dose verification Med. Phys. 29 1269
Nill S, Tucking T, Munter M and Oelfke U 2002 IMRT with MLCS of different leaf widths:
a comparison of achievable dose distributions Radiother. Oncol. 64 (Suppl. 1) S1078
Nioutsikou E, Bedford J L, Webb S and Christian J 2004 Segmentation of IMRT plans
for radical lung radiotherapy delivery with the step-and-shoot technique Radiother.
Oncol. 31 892901
Nishioka T, Shiga T, Shirato H, Tsukamoto E, Tsuchiya K, Kato T, Ohmori K, Yamazaki
A, Aoyama H, Hashimoto S, Chang T-C and Mikyasaka K 2002 Image fusion
between 18 FDG-PET and MRI/CT for radiotherapy planning of oropharyngeal and
nasopharyngeal carcinomas Int. J. Radiat. Oncol. Biol. Phys. 53 10517
Nishioka T, Shirato H, Karo T, Watanabe Y, Yamazaki A, Ohmori K, Aoyama H, Shiga
T, Tsukamoto E, Hashimoto S, Tsuchiya K and Miyasaka K 2000 Impact of 18 FDGPET and CT/MRI image fusion in radiotherapy planning of head-and-neck tumours
Proc. 42nd Annual ASTRO Meeting. Int. J. Radiat. Oncol. Biol. Phys. 48 (Suppl.) 260
Novotny J Jnr, Dvorak P, Spevacek V, Tintera J, Novotny J, Cechak T and Liscak R 2002
Polymer-gel dosimetry as a tool for quality control of the stereotatic radiosurgery
procedure Radiother. Oncol. 64 (Suppl. 1) S44
Nusslin F 2001 IMRT and beyondperspective in radiation therapy Phys. Medica. 17
1634
Nusslin F, Alber M, Baum C, Birkner M and Buck D 2003 Early steps toward IGRT:
Development of imaging tools for IMRT treatment planning radiotherapy Phys.
Medica. 19 45
Nusslin F, Alber M, Birkner M and Fippel M 2003 Hyperiona novel Monte-Carlo
IMRT treatment planning program based on isoeffect dose optimisation Proc. World
Nutting C 2003 Intensity-modulated radiotherapy (IMRT): the most important advance in
radiotherapy since the linear accelerator? Brit. J. Radiol. 76 673
Nutting C and Dearnaley D 2001 Intensity-modulated radiation therapy (IMRT) for
tumours of the head and neck, pelvis and thorax: pre-clinical evaluation and
implementation Compendium of the 3rd IMRT Winter School, Heidelberg (London,
Jan 26/27) pp C2.1C2.3
2002 Prostate cancerthe Royal Marsden conformal experience Normal Tissue
Reactions in Radiotherapy and Oncology ed W Dorr, R Engenhart-Cabillic and
J S Zimmermann (Basel: Karger) pp 1969
Nutting C and Harrington K 2004 (in regards to Dr Glatstein 2003 Int. J. Radiat. Oncol.
Biol. Phys. 55 5612 and Dr Amols 203 Int. J. Radiat. Oncol. Biol. Phys. 56 1507)
References
433
Nutting C M, Rowbottom C, Cosgrove V P, Convery D J, Webb S and Dearnaley D P

2000 Evaluation of intensity modulated radiotherapy for tumours of the pelvis, head
and neck, and thorax: Optimisation and clinical implementation Proc. 42nd Annual
OBrien P, Woo M and Nico A 2002 Beam energy selection for intensity modulated
radiation therapy Radiother. Oncol. 64 (Suppl. 1) S212
ODaniel J C 2001 Comparison of 3D conformal and intensity-modulated radiation therapy
for treatment of prostate cancer Med. Phys. 28 1292
ODaniel J, Dong L, Kuban D A, Liu H, Schechter N, Tucker S L and Rosen I 2004 The
delivery of IMRT with a single physical modulator for multiple fields: a feasibility
study for paranasal sinus cancer Int. J. Radiat. Oncol. Biol. Phys. 58 87687
ODaniel J, Dong L, Rosen I and Kuban D 2002 Comparison of planning quality and
delivery efficiency of two commercial IMRT planning systems Med. Phys. 29 1283
Oelfke U 2001 Delivery techniques of IMRT: MLCs Radiother. Oncol. 61 (Suppl. 1) S1
Oelfke U, Rhein B and Haring P 2001 Dosimetric procedures and problems of IMRT Proc.
Topical Meeting on Medical Radiation Physics and Engineering, Lisbon, Portugal,
Nov 2022, 2000 Phys. Medica. 17 (Suppl. 4) 302
Oelfke U, Siemer N and Nill S 2002 Optimization of IMRT treatment plans with EUDbased objective functions Radiother. Oncol. 64 (Suppl. 1) S51
Oh Y, Ji Y, Kim K, Cho M, Kim J, Jeong D, Shin K and Kim J 2003 The realization of
intensity-modulated radiation therapy using the liquid shielding material Proc. World
Olch A 2001 Clinical implementation of IMRT in a small US clinic Radiother. Oncol. 61
(Suppl. 1) S74
Oldham M and Kim L 2004 Optical-CT gel-dosimetry 2: optical artifacts and geometrical
distortion Med. Phys. 31 1093104
Oldham M, McJury M, Baustert I B, Webb S and Leach M O 1998 Improving calibration
accuracy in gel-dosimetry Phys. Med. Biol. 43 270920
Oldham M, Siewerdsen J H, Kumar S, Wong J and Jaffray D A 2003 Optical-CT geldosimetry 1: basic investigations Med. Phys. 30 62334
Oldham M, Siewerdsen J H, Shetty A and Jaffray D A 2001 High resolution gel-dosimetry
by optical and MR scanning Med. Phys. 28 143645
Oliver L, Phillips R, Vial P, Hunt P and Mallik A 2002 Conformal versus IMRT treatment
plans Radiother. Oncol. 64 (Suppl. 1) S78
Olivera G, Fitchard E, Reckwerdt P, Ruchala K and Mackie R 2000b Modifying a plan
delivery without re-optimization to account for patient offset in tomotherapy Proc.
paper MO-EBR-01
Olivera G, Kapatoes J, Mackie T, Lu W, Reckwerdt P, Jeraj R, Balog J and Ruchala K 2002
Use of helical tomotherapy optimization from very simple to very complex treatment
planning Radiother. Oncol. 64 (Suppl. 1) S77
Olivera G H, Mackie T R, Ruchala K J, Kapatoes J M, Reckwerdt P J, Jeraj R, Balog J P,
Lu W, Smilowitz and Keller H 2001 Tomotherapy: a tool for image guided adaptive
radiotherapy Radiother. Oncol. 61 (Suppl. 1) S24
Olivera G, Reckwerdt P, Shepard D M and Mackie R 2000a Large-scale helical
tomotherapy optimization: Four clinical case studies Proc. World Congress of
Medical Physics and the AAPM Annual Congress, August 2000 paper FR-B309-04
434
References
Onishi H, Kuriyama K, Komiyama T, Ueki J and Araki T 2001 A new irradiation system
for lung cancer: patients self-breath-hold and self-switching radiationbeam on and
off without any respiratory monitoring devices Int. J. Radiat. Oncol. Biol. Phys. 51
(Suppl. 1) 26
Orton N and Tome 2003 Dosimetric impct of daily shifts on prostate IMRT dose
distributions Med. Phys. 30 1339
Otto K, Alkins R and Clark B G 2003 Limitations of IMRT planning with low resolution
pencil beam kernels Radiother. Oncol. 68 (Suppl. 1) S43
Otto K and Clark B G 2002 Enhancement of IMRT delivery through MLC rotation Phys.
Med. Biol. 47 39974017
Ozhasoglu C, Christie N, Luketich J, Burton S, Gerszten P, Wu A and Kalnicki S 2003
Target immobilization with natural breath-holding: Possible applications to IMRT
and body radiosurgery Med. Phys. 30 1338
Ozhasoglu C, Murphy M J and Adler J R 2001 Real time compensation of breathing motion
in radiotherapy Med. Phys. 28 1295
Paliwal B, Balog J, Smilowitz J, Das R and Mackie T 2002a Commissioning dosimetry for
a UW clinical tomotherapy unit Med. Phys. 29 121112
Paliwal B, Balog J, Tome W, Das R, Richardson S, Mackie R and Mehta S 2002b
Tomotherapy: initial University of Wisconsin experience Radiother. Oncol. 64
(Suppl. 1) S22
Paliwal B, Balog J, Tome W, Richardson S and Mackie T 2001 Dosimetry verification for
IMRT/Tomotherapy Radiother. Oncol. 61 (Suppl. 1) S16
Paliwal B R, Brezovich I A and Hendee W R 2004 IMRT may be used to excess because
of its higher reimbursement from medicare Med. Phys. 31 13
Paliwal B Tome W, Richardson S and Mackie T R 2000 A spiral phantom for IMRT and
tomotherapy treatment delivery verification Med. Phys. 27 25037
Palta J R and Mackie T R (eds) 2003 IMRTthe State of the Art: AAPM Medical Physics
Monograph Number 29 (Madison, WI: Medical Physics Publishing)
Papatheodorou S, Castellanos M, Zefkili S, Gaboriaud G and Rosenwald J 2000b Dynamic
collimation for missing tissue compensation Proc. World Congress of Medical
Physics and the AAPM Annual Congress, August 2000 paper MO-CXH-02
Papatheodorou S, Rosenwald J-C, Zefkili S, Murillo M-C, Drouard J and Gaboriaud
G 2000a Dose calculation and verification of intensity modulation generated by
dynamic multileaf collimators Med. Phys. 27 96071
Papanikolaou N 2001 Dose calculation algorithms in the IMRT era Radiother. Oncol. 61
(Suppl. 1) S12
Papanikolaou N, Gutierrez L, Yan Y, Wu C, Penagaricano J and Ratanatharathorn V
2003 The effect of tissue inhomogeneity in dose based and biologically based IMRT
Papanikolaou N, Stathakis S, Tsougo Y and Kappas C 2001 The effect of tissue
inhomogeneity on IMRT Med. Phys. 28 1207
Papanikolaou N, Yan Y, Penagaricano J and Ratanatharathorn V 2002 Application of IMRT
for the treatment of medulloblastoma Med. Phys. 29 1215
Papiez L, Thai V and Langer M 2001 Optimal step-and-shoot segmentation algorithm for
unidirectional motion of leaves Med. Phys. 28 1204
Parker B, Shiu A and Comiskey R 2000 Comparison of dose distributions from geometric
conformation and intensity modulation for intracranial lesions Proc. World Congress
References
435
of Medical Physics and the AAPM Annual Congress, August 2000 paper MO-BBR05
Parker C C, Damyanovich A, Haycocks T, Haider M, Bayley A and Catton C N
2003 Magnetic resonance imaging in the radiation treatment planning of localized
prostate cancer using intra-prostatic fiducial markers for computed tomography coregistration Radiother. Oncol. 66 21724
Parker W and Brodeur M 2002 Measurement of 3D dose distributions and DVH of IMRT
treatment plans using MR based gel dosimetry Med. Phys. 29 12634
Parker W, Hristov D, Moftah B, Vuong T, Tsien C and Podgorsak E 2000 Intensity
modulated radiation therapy for thyroid malignancies Proc. World Congress of
Medical Physics and the AAPM Annual Congress, August 2000 paper TU-GBR-02
Parsai H, Cho P S, Phillips M H, Giansiracusa R S and Axen D 2003 Random and
systematic beam modulator errors in dynamic intensity modulated radiotherapy Phys.
Med. Biol. 48 110921
Parsai H, Phillips M H and Cho P S 2001a Dosimetry verifications of dynamic IMRT: An
in-vivo study in canine subjects Med. Phys. 28 1208
2001b Verification of dynamic intensity-modulated beam deliveries in canine subjects
Med. Phys. 28 2198208
Partridge M 2000 Reconstruction of megavoltage photon spectra from electronic portal
imager derived transmission measurements Phys. Med. Biol. 45 N11531
Partridge M, Aldridge S, Donovan E and Evans P M 2001a An intercomparison of IMRT
delivery techniques: a case study for breast treatment Phys. Med. Biol. 46 N17585
Partridge M, Christian J A, Flux G, McNair H A, Cronin B, Cook G and Courbon F 2003
Accurate co-registration of CT and SPECT perfusion images for lung radiotherapy
planning Radiother. Oncol. 68 (Suppl. 1) S67
Partridge M, Ebert M and Hesse B-M 2002 IMRT verification by three-dimensional dose
reconstruction from portal beam measurements Med. Phys. 29 184758
Partridge M, Evans P M, van Herk M, Ploeger L S, Budgell G J and James H V 2000a
Leaf position verification during dynamic beam delivery: a comparison of three
applications using electronic portal imaging Med. Phys. 27 16019
Partridge M and Hesse B M 2002 IMRT dose verification in-vivo in 3D Radiother. Oncol.
64 (Suppl. 1) S123
Partridge M, Hesse B M, Groh B, Ebert M and Bortfeld T 2001b Verification of 3D dose
delivery at treatment time Radiother. Oncol. 61 (Suppl. 1) S4
Partridge M, Symonds-Tayler J and Evans P 2000b Dynamic MLC verification with portal
imaging Radiother. Oncol. 56 (Suppl. 1) S89
Partridge M, Symonds-Tayler J R N and Evans P M 2000c IMRT verification with a
camera-based electronic portal imaging system Phys. Med. Biol. 45 N18396
Paskalev K, Ma C, Jacob R and Price R 2003 Daily target localization based on 3D image
correlation Med. Phys. 30 1337
Paskalev K, Ma C-M, Jacob R, Price R, McNeeley S, Wang L, Movsas B and Pollack A
2004 Daily target localization for prostate patients based on 3D image correlation
Pastyr O, Echner G, Hartmann G, Richter J and Schlegel W 2001 Dynamic edge focussing:
A new MLC-design to deliver IMRT with a double focussing high precision multileaf-collimator Radiother. Oncol. 61 (Suppl. 1) S24
436
References
Pasma K L, Vieira S C, de Langen M, Dirkx M L P, Storchi P and Heijmen B J M

2001 Dosimetric verification of dynamic multileaf collimation using a fluoroscopic
electronic portal imaging device (EPID) Radiother. Oncol. 61 (Suppl. 1) S16
Patel R R, Tome W A, Tannehill S P, Harari P M, Paliwal B R and Mehta M P 2002
Optically guided intensity modulated radiotherapy for the head and neck Radiother.
Paulino A C and Johnstone P A S 2004 FDG-PET in radiotherapy treatment planning:
Pandoras box? Int. J. Radiat. Oncol. Biol. Phys. 59 45
Paulsen F, Alber M, Eschmann S M, Plasswilm L, Budach W, Machulla H J, Nusslin F,
Bares R and Bamberg M 2002 Using PET imaging and IMRT to generate a smart
hypoxia boost for head and neck tumours Radiother. Oncol. 64 (Suppl. 1) S92
Pavel L, Rowbottom C G, Hector C, Partridge M and Bortfeld T 2001 Estimation of the
effects of movements on dose distribution for prostate patients Radiother. Oncol. 61
(Suppl. 1) S36
Pavel-Mititean L M, Rowbottom C G, Hector C L, Partridge M, Bortfeld T and Schlegel
W 2004 A geometrical model for evaluating the effects of inter-fraction rectal motion
during prostate radiotherapy Phys. Med. Biol. 49 261329
Pawlicki T, Jiang S B, Li J, Deng J and Ma C M 2000a Including setup uncertainty in
Monte Carlo dose calculation for IMRT Proc. World Congress of Medical Physics
and the AAPM Annual Congress, August 2000 paper MO-B309-02
2000b The role of set-up uncertainty in intensity modulated treatment planning Proc.
Pawlicki T, Luxton G, Le Q-T, Findley D and Ma C-M 2004 Lens dose in MLC-based
IMRT treatments of the head and neck Int. J. Radiat. Oncol. Biol. Phys. 59 2939
Pelizzari C A 2003 Medical imaging in IMRT planning IMRTthe State of the Art: AAPM
Peltola S 2003 Intensity modulation radiotherapy using compensatorsdose calculations
using CADPLAN HELIOS dose planning Proc. World Congress in Medical Physics
(Sydney) CDROM and website only
Pemler P, Besserer J, Lombriser N, Pescia R and Schneider U 2001 Influence of respirationinduced organ motion on dose distributions in treatments using enhanced dynamic
wedges Med. Phys. 28 223440
Perrin B, Budgell G J and Mackay R I 2003 A tool for quantitative analysis of IMRT
verification measurements Clinical Oncol. 15 (Suppl. 2) S32
Petersen J B, Spejlborg H, Hansen A T, Thomsen M S, Tanderup K, Vestergaard A, Heyer
M, Lindegaard J C, Ulso N and Grau C 2002 Implementation of intensity modulated
radiotherapy in Aarhus, Denmark Radiother. Oncol. 64 (Suppl. 1) S216
Phillips M 2002 IMRT of lung cancer Med. Phys. 29 12589
Phillips M H, Cho P S and Parsaei H 2001a Validation of an IMRT Program from
A(lgorithm) to Z(ooRadiotherapy) Radiother. Oncol. 61 (Suppl. 1) S1
Phillips M H, Cho P S, Parsaei H, Kippenes H and Gavin P 2000 Verification of dynamic
IMRT with in vivo and phantom dosimetry Radiother. Oncol. 56 (Suppl. 1) S89
Phillips M H, Meyer J, Cho P S, Kalet I J and Doctor J N 2004 Bayesian decision making
applied to IMRT Proc. 14th Int. Conf. on the Use of Computers in Radiation Therapy
(Seoul, May) pp 108-111
Phillips M, Parsaei H and Cho P 2001b Modelling dynamic IMRT dose distributions Med.
Phys. 28 1293
References
437
Pickett B, Pirzkall A, Kurhanawicz J, Verhey L and Roach M 2000b Radiosurgical intensity

modulated radiotherapy for prostate cancer Proc. 42nd Annual ASTRO Meeting. Int.
J. Radiat. Oncol. Biol. Phys. 48 (Suppl.) 138
Pickett B, Xia P, Woods P, Kurhanawicz J and Roach M 2000a Most dramatic magnetic
resonance spectroscopy imaging changes correlate with high dose regions following
intensity modulated radiotherapy for prostate cancer Radiother. Oncol. 56 (Suppl. 1)
S106
Pignoli E, Serretiello S, Somigliana A, Zonca G, Pellegrini R, Mongioj V and Marchesini
R 2000 Dosimetric verification of a commercial 3D treatment planning system for
conformal radiotherapy with a dynamic multileaf collimator Phys. Med. Biol. 45
N7784
Pirzkall A, Carol M, Lohr F, Hoess A, Wannenmacher M and Debus J 2000 Comparison
of intensity-modulated radiotherapy with conventional conformal radiotherapy for
complex-shaped tumours Int. J. Radiat. Oncol. Biol. Phys. 48 137180
Pirzkall A, Carol M, Pickett B, Roach III M, Verhey L 2000a The effect of beam energy
and number of fields on photon-based IMRT for deep seated targets Proc. World
MO-EBR-10
Pirzkall A. Carol M P, Pickett B, Xia P, Roach M and Verhey L J 2002 The effect of beam
energy and number of fields on photon-based IMRT for deep-seated targets Int. J.
Pirzkall A, Debus J, Haering P, Rhein B, Grosser K, Hoss A and Wannenmacher M 2003
Intensity modulated radiotherapy (IMRT) for recurrent, residual, or untreated skullbase meningiomas: preliminary clinical experience Int. J. Radiat. Oncol. Biol. Phys.
55 36272
Plasswilm L, Alber M, Paulsen F, Bakai A, Birkner M, Glocker S, Christ G, Buck
C, Nusslin F and Bamberg M 2002 First clinical experience with image-guided
intensity modulated radiation treatment of prostate cancer patients Radiother. Oncol.
64 (Suppl. 1) S271
Ploeger L S, Smitsmans M H P, Gilhuijs K G A and van Herk M 2002 A method
for geometrical verification of dynamic intensity modulated radiotherapy using a
scanning electronic portal imaging device Med. Phys. 29 10719
Pollack A and Price R A 2003 IMRT for prostate cancer IMRTthe State of the Art: AAPM
Pollack A, Zagars G K, Antolak J A, Kuban D A and Rosen I I 2002b In response to Drs.
Kagan and Schulz Int. J. Radiat. Oncol. Biol. Phys. 55 11512
Pollack A, Zagars G K, Smith L G, Lee J, von Eschenbach A C, Antotak J A, Starkschall
G and Rosen I 2000 Preliminary results of a randomized radiotherapy dose escalation
study comparing 70 Gy with 78 Gy for prostate cancer J. Clin. Oncol. 18 390411
Pollack A, Zagars G K, Starkshall G, Antotak J A, Lee J, Huang E, von Eschenbach A C,
Kuban D A and Rosen I 2002a Prostate cancer radiation dose response: results of the
M D Anderson Phase-3 trial Int. J. Radiat. Oncol. Biol. Phys. 53 1097105
Popple R, Prellop P, Spencer S, Fiveash J, Duan J and Brezovich I 2003 Simultaneous
optimisation of sequential IMRT plans Radiother. Oncol. 68 (Suppl. 1) S76
Portelance L, Chao K S C, Grigsby P W, Bennet H and Low D 2001 Intensity-modulated
radiation therapy (IMRT) reduces small bowel, rectum and bladder doses in patients
438
References
with cervical cancer receiving pelvic and para-aortic irradiation Int. J. Radiat. Oncol.
Biol. Phys. 51 2616
Potter L, Chang S and Cullip T 2003a Intensity modulation via a 4-bank micro-MLC
system Med. Phys. 30 1348
Potter L D, Chang S X, Cullip T J and Siochi A C 2002 A quality and efficiency analysis
of the IMFAST segmentation algorithm in head and neck step and shoot IMRT
treatments Med. Phys. 29 27583
Potters L, Steinberg M, Wallner P and Hevezi J 2003b How one defines intensitymodulated radiation therapy Int. J. Radiat. Oncol. Biol. Phys. 56 60910
Pouliot J, Aubin M, Verhey L, Bani-Hashemi A, Mitschke M, Hernandez P and Hughes
J 2002 Low dose megavoltage CT cone beam reconstruction for patient alignment
Med. Phys. 29 1241
Pouliot J, Xia P, Aubin M, Verhey L, Langen K, Bani-Hashemi A, Svatos M, Ghelmansarai
F and Mitchke M 2003 Dose-guided radiation therapy using low-dose megavoltage
cone-beam CT Med. Phys. 30 13378
Poynter A 2003 Intensity modulated radiotherapya clinical experience in a small
oncology centre RAD Magazine 29 (334) 356
Price P 2002 Functional imaging to define radiotherapy target volumesthe future
Price R, Ma C, McNeeley S, Chen L, Fourkal E, Li J, Horwitz E and Pollack A 2003 IMRT
at Fox Chase Cancer Center Radiother. Oncol. 68 (Suppl. 1) S63
Price R, Murphy S, McNeeley S and Ma C 2002 A method for increased dose conformity
and segment reduction for SMLC delivered IMRT treatment of the prostate Med.
Phys. 29 1257
Price R A, Murphy S, McNeeley S W, Ma C-M, Horwitz E, Movsas B, Raben A and
Pollack A 2003 A method for increased dose conformity and segment reduction for
sMLC delivered IMRT treatment of the prostate Int. J. Radiat. Oncol. Biol. Phys. 57
84352
Pugachev A B, Boyer A L and Xing L 2000a Beam orientation optimization in intensitymodulated radiation treatment planning Med. Phys. 27 123845
Pugachev A B, Hancock S, Le Q T, Boyer A L and Xing L 2000b Pseudo beams-eye-view
(BEV) applied to beam orientation selection for IMRT Proc. 42nd Annual ASTRO
Pugachev A, Li J, Boyer A and Xing L 2000c Role of non-coplanar beams in IMRT Proc.
paper MO-EBR-05
Pugachev A, Li M, Boyer A L, Hancock S L, Le Q-T, Donaldson S S, and Xing L 2001
Role of beam orientation optimization in intensity-modulated radiation therapy Int.
Pugachev A and Xing L 2001a Two stage beam orientation selection method for IMRT
Med. Phys. 28 1204
2001b Computer assisted selection of coplanar beam orientations in intensitymodulated radiation therapy Phys. Med. Biol. 46 246776
2001c Pseudo beams-eye-view as applied to beam orientation selection in intensitymodulated radiation therapy Int. J. Radiat. Oncol. Biol. Phys. 51 136170
2001d Computer assisted selection of beam energy and orientation in IMRT Int. J.
References
439
2002a Examination of the degree of validity of class-solution approach for IMRT

Med. Phys. 29 1284
2002b Incorporating prior knowledge into beam orientation optimisation in IMRT.
Int. J. Radiol. Oncol. Biol. Phys. 54 156574
Purdy J A and Michalski J M 2001 Does the evidence support the enthusiasm over 3D
conformal radiation therapy and dose escalation in the treatment of prostate cancer?
2002 In response to Drs Levitt and Khan Int. J. Rad. Onc. Biol. Phys. 53 1085
Que W 1999 Comparison of algorithms for multileaf collimator field segmentation. Med.
Phys. 26 23906
Que W, Kung J and Dai J 2004 Tongue-and-groove effect in intensity modulated
radiotherapy with static multileaf collimator fields Phys. Med. Biol. 49 399405
Raaymakers B W, Lagendijk J J W, van der Heide U A, Overweg J, Brown K, Topolnjak R,
Dehnad H, Jurgenliemk-Schulz I M, Welleweerd J and Bakker C J G 2004 Integrating
a MRI scanner with a radiotherapy accelerator: a new concept of precise on line
radiotherapy guidance and treatment monitoring Proc. 14th Int. Conf. on the Use of
Raaymakers B W, Lagendijk J J W, van der Heide U A and Topolnak R 2003 The
development of a Linac-MRI combination for on-line position verification Phys.
Medica. 19 456
Raben A, Yang J, Murphy S, Stapleton T, Chen H, Grebler A, Geltzeller J, Simmons
D and Gollamudi S 2002 Feasibility and safety of high dose intensity modulated
radiotherapy (IMRT) for localized prostate cancer in a community practice:
preliminary results on acute toxicity Radiother. Oncol. 64 (Suppl. 1) S277
Radford D L, Followill D and Hanson W 2000 Design of an anthropomorphic intensity
modulated radiation therapy quality assurance phantom Proc. World Congress of
Medical Physics and the AAPM Annual Congress, August 2000 paper MO-CXH-01
2001 A standard method of quality assurance for intensity modulated radiation
therapy of the prostate Med. Phys. 28 1211
Ramaseshan R, Cho Y, Purdie T, Alasti H, Wong R, Kohli K, Chan B, Heaton R and Islam
M 2003 Breathing motion in lung: is external surface surrogate for internal structure?
Ramsey C 2002 The effects of respiration motion on dynamic IMRT Med. Phys. 29 12601
Ramsey C, Dube S and Hendee W R 2003 It is necessary to validate each individual IMRT
treatment plan before delivery Med. Phys. 30 22713
Ramsey C R, Large C, Scaperoth D, Arwood D and McDonald W 2001b The dosimetric
consequences of intra-fraction prostate motion in intensity modulation Int. J. Radiat.
Ramsey C R, Scaperoth D and Arwood D 2000 Clinical experience with a commercial
respiratory gating system Proc. 42nd Annual ASTRO Meeting. Int. J. Radiat. Oncol.
Ramsey C R, Spencer K M, Alhakeem R and Oliver A L 2001a Leaf position error during
conformal dynamic arc and intensity modulated arc treatments Med. Phys. 28 6772
Remouchamps V, Claus F, Vermael S, de Gersem W, Vakaet L and de Neve W 2000
Intensity modulated radiation therapy for large breast: physical wedges versus
segmented tissue compensation Radiother. Oncol. 56 (Suppl. 1) S107
Remouchamps V M, Vicini F A, Martinez A A, Sharpe M B, Kestin L L and Wong
J W 2002 Reductions in heart and lung doses using deep inspiration breath hold
440
References
with active breathing control and IMRT for patients treated with locoregional breast
irradiation Radiother. Oncol. 64 (Suppl. 1) S139
Remouchamps V M, Vicini F A, Sharpe M B, Kestin L L, Martinez A A and Wong J W
2003 Significant reductions in heart and lung doses using deep inspiration breath hold
with active breathing control and intensity-modulated radiation therapy for patients
treated with locoregional breast irradiation Int. J. Radiat. Oncol. Biol. Phys. 55 392
406
Renner W D, Sarfaraz M, Earl M A and Yu C X 2003 A dose delivery verification
method for conventional and intensity-modulated radiation therapy using measured
field fluence distributions Med. Phys. 30 29963005
Rhein B and Haring P 2001 The IMRT phantom verification procedure in Heidelberg
Richardson S L , Tome W A, Orton N P, McNutt T R and Paliwal B R 2003 IMRT delivery
verification using a spiral phantom Med. Phys. 30 25538
Richter J, Hadinger U, Kolbl O, Neumann M and Flentje M 2000 IMRT versus quasi
dynamic rotation using MLC Radiother. Oncol. 56 (Suppl. 1) S195
Rietzel E, Chen G T Y, Adams J A, Hernandez A H and Willett C G 2004 Treatment
planning including respiratory motion based on 4D computed tomography data Proc.
14th Int. Conf. on the Use of Computers in Radiation Therapy (Seoul, May) pp 27-31
Rietzel E, Doppke K, Pan T, Choi N, Willett C and Chen G 2003 4D computer tomography
for radiation therapy Med. Phys. 30 13656
Rivard M, Tsai J-S and Engler M 2002 Implementation of inversely-planned, intensity
modulated radiation therapy using doubly-dynamic MLC Med. Phys. 29 1268
Rock L, McClean B, McCavana P and Buckney S 2001 Comparison of calculated
and measured data for small MLC generated fields defined on Varian and Elekta
accelerators Radiother. Oncol. 61 (Suppl. 1) S24
Rodebaugh R, Crownover R, Meziane M, Rice T, Graham R, Mehta A, Bodduluri M,
Glosser G, Wang J, Chen Q S, Macklis R and Weinhous M 2001a Chasing lung
lesions with the Cyberknife Med. Phys. 28 1219
Rodebaugh R F, Crownover R L, Weinhous M S, Meziane M, Graham R I, Glosser G,
Wang J, Bodduluri M, Whyte R I and Rice T W 2001b The accuracy of tracking lung
tumors with the Cyberknife. Int. J. Radiat. Oncol. Biol. Phys. 51 (Suppl. 1) P245
Roeske J, Lujan A, Rotmensch J and Mundt A 2000a A feasibility study of IMRT for the
treatment of cervical cancer patients unable to receive intracavitary brachytherapy
2000 paper MO-EBR-07
Roeske J C, Lujan A, Rotmensch J, Waggoner S E, Yamada D and Mundt A J 2000b
Intensity-modulated whole pelvic radiation therapy in patients with gynaecologic
malignancies Int. J. Radiat. Oncol. Biol. Phys. 48 161321
Rohlfing T, Maurer C R, ODell W G and Zhong J 2004 Modeling liver motion and
deformation during the respiratory cycle using intensity-based nonrigid registration
of gated MR images Med. Phys. 31 42732
Romeijn H E, Ahuja R K, Dempsey J F, Kumar A and Li J G 2003 A novel linear
programming approach to fluence map optimization for intensity modulated radiation
therapy treatment planning Phys. Med. Biol. 48 352142
Rose A, Short R, Hoban P and Biggs D 2001 A stereotactically located phantom for
verification of IMRT patient treatments Radiother. Oncol. 61 (Suppl. 1) S109
References
441
Rosello J, Ferrando I, Gonzalez Lopez A, Perez Calatayud J, Lopez Torrecilla J,

Dualde Beltran D, Gonzalez Sanchis A and Hernandez Mechancoses A 2000
Implementation of intensity modulated radiotherapy (IMRT) with HELAX-TMS
inverse dose planning system and PRIMUS-SIEMENS linear accelerator: problems
and solutions Radiother. Oncol. 56 (Suppl. 1) S196
Rosello J, Gonzalez A, Sanchez-Doblado F, Leal A, Arrans R, Nunez L and Brualla L 2001
Weak points of an inverse treatment planning system Radiother. Oncol. 61 (Suppl. 1)
S110
Rosu M, Dawson L A, Balter J M, McShan D L, Lawrence T S and Ten Haken R K 2003
Alterations in normal liver doses due to organ motion Int. J. Radiat. Oncol. Biol.
Phys. 57 14729
Rousseau V, Grandjean P, Diaz J C, Kafrouni H, Berre F and Lefkopoulos D 2002
Comparison between two inverse methods for IMRT Radiother. Oncol. 64 (Suppl.
1) S217
Rowbottom C G 1998 Optimisation of beam orientations in conformal radiotherapy
treatment planning PhD Thesis London University
Rowbottom C G, Nutting C M and Webb S 2001 Beam optimization of intensity-modulated
radiotherapy: clinical application to parotid gland tumours Radiother. Oncol. 59 169
77
Ruchala K, Olivera G, Forrest L, Mehta M, Kapatoes J, Welsh J and Mackie T 2002c
Megavoltage CT for image-guided radiotherapy Radiother. Oncol. 64 (Suppl. 1) S12
Ruchala K, Olivera G, Kapatoes J, Balog J, Welsh J, Reckwerdt P, Jeraj R, Lu W and
Mackie T 2003 On-board CT image quality for image-guided radiotherapy Proc.
Ruchala K, Olivera G, Kapatoes J, Jeraj R, Reckwerdt P and Lu W 2002b Multimargin optimization with daily selection (MMODS) for image-guided radiotherapy
Ruchala K, Olivera G, Kapatoes J, Lu W, Reckwerdt P, Jeraj R and Mackie T 2002a A
solution for the Cassandra Complex of image-guided radiotherapy Med. Phys. 29
1240
Ruchala K J, Olivera G H, Kapatoes J M, Reckwerdt P J and Mackie T R 2002d Methods of
improving limited field-of-view radiotherapy reconstructions using imperfect a priori
images Med. Phys. 29 2590605
Ruchala K J, Olivera G H, Kapatoes J M, Schloesser E A, Reckwerdt P J and Mackie T
R 2000a Megavoltage CT image reconstruction during tomotherapy treatments Phys.
Med. Biol. 45 354562
2000b Megavoltage CT imaging as a by-product of multileaf collimator leakage Phys.
Med. Biol. 45 N6170
Ryu J, Winter K, Michalski J, Purdy J A, Markoe A M, Earle J D, Perez C A, Roach
M III, Sandler H M, Pollack A and Cox J D 2001 Interim report of toxicity from
3D conformal radiation therapy for prostate cancer on 3DOG/RTOG 9406, Level III
(79.2 Gy) ASCO Proc. abstract 725, J. Clin. Oncol.
Salter B 2001 NOMOS Peacock IMRT utilizing the Beak post collimation device Medical
Dosim. 26 3745
Salter B J, Fuss M, Vollmer D G, Sadeghi A, Bogaev C A, Cheek D A, Herman T
S and Hevezi J M 2001 The talon removable head frame system for stereotactic
radiosurgery/radiotherapy: measurement of the repositioning accuracy Int. J. Radiat.
442
References
Salter B, Voeltz L and Fuss M 2003 Characterizing the dosimetric impact of BAT
ultrasound alignment Med. Phys. 30 1339
Samuelsson A and Johansson K 2003 Intensity modulated radiotherapy treatment planning
for dynamic multileaf collimator delivery: influence of different parameters on dose
distributions Radiother. Oncol. 66 1928
Samuelsson A, McShan D L, Ten Haken R K, Eisbruch A, Johansson K A and Fraass B
A 2001 Set-up uncertainties in optimised IMRT plans for the head and neck: plan
degradation and correction Radiother. Oncol. 61 (Suppl. 1) S47
Samuelsson A, Mercke C and Johansson K-A 2003 Systematic set-up errors for IMRT in
the head and neck region: effect on dose distribution Radiother. Oncol. 66 30311
Sanchez Doblado F, Capote R, Leal A, Lagares J, Arrans R and Hartmann G 2003
Absolute dosimetry of narrow irregular centred IMRT beamlet Proc. World Congress
Sanchez-Doblado F, Leal A, Arrans R, Rosello J, Carrasco E and Perucha M 2001
Automated assessment of IMRT treatments Radiother. Oncol. 61 (Suppl. 1) S101
Sanchez-Nieto B 2000 Practical use of TCP in treatment plan optimization Radiother.
Sanford R, Al-Ghazi M, Chung H and Yakoob R 2000 Dosimetric validation of clinical
serial tomotherapy delivery Proc. World Congress of Medical Physics and the AAPM
Annual Congress, August 2000 paper WE-E313-01
Saw C, Ayyangar K, Zhen W, Shen B and Enke C 2001 Assessment of phantom size for
Saw C B, Li S, Ayyangar K M, Yoe-Sein M, Pillai S, Enke C A and Celi J C 2003
Dose linearity and uniformity of a linear accelerator designed for implementation of
multileaf collimation system-based intensity modulated radiation therapy Med. Phys.
30 22536
Sawada A, Yoda K, Kokubo M, Nagata Y and Hiraoka M 2002 Non-invasive respiratory
gated radiation treatment system based on a 3D ultrasound device and a 3D digital
localizer Med. Phys. 29 1310
2004 A technique for noninvasive respiratory gated radiation treatment system based
on a real time 3D ultrasound image correlation: a phantom study Med. Phys. 31
24550
Schaefer M, Munter M W, Thilmann C, Sterzing F, Haering P, Combs S E and Debus J 2004
Influence of intra-fractional breathing movement in step-and-shoot IMRT Phys. Med.
Biol. 49 N1759
Schaly B, Kempe J A, Bauman G S, Battista J J and van Dyke J 2004 Tracking the dose
distribution in radiation therapy by accounting for variable anatomy Phys. Med. Biol.
49 791805
Schefter T, Kavanagh BH, Newell S and McCourt S 2002 Solid-phase compensator
- versus multi-leaf collimator-based intensity-modulated radiotherapy for cervix
cancer: dosimetric comparison, resource utilization, and financial considerations
Schlegel W 2003 Implementation of IMRT Clinical Oncol. 15 (Suppl. 2) S5
Schlegel W, Oelfke U, Thieke C, Bortfeld T and Kufer K H 2003 Inverse planning for
IMRT with Konrad Phys. Medica. 19 53
Schmuecking M, Plichta K, Lopatta E C, Przetak C, Leonhardi J, Gottschild D, Wendt T
G and Baum R P 2000 Image fusion of F-18 FDG PET and CTis there a role in
References
443
3D-radiation treatment planning of non-small cell lung cancer? Proc. 42nd Annual
ASTRO Meeting. Int. J. Radiat. Oncol. Biol. Phys. 48 130
Schmidt W F O, Krauss H, Poljanc K, Buechler F and Wirth P 2001 IMRT in routine:
questions to be answered Radiother. Oncol. 61 (Suppl. 1) S58
Schmidt W, Nespor W, Wirth P, Beuchier F and Hawliczek R 2000 IMRT in routine Experiences from a feasibility study Proc. World Congress of Medical Physics and
the AAPM Annual Congress, August 2000 paper MO-CXH-21
Schreibmann E, Lahanas M, Xing L and Baltas D 2004c Multiobjective evolutionary
optimization of the number of beams, their orientations and weights for intensitymodulated radiation therapy Phys. Med. Biol. 49 74770
Schreibmann E, Theodorou K, Kappas C and Xing L 2004a A software package for dose
visualization in IMRT Proc. 14th Int. Conf. on the Use of Computers in Radiation
Schreibmann E and Xing L 2004b Score-based beam selection in IMRT with inclusion of
EUD constraints Proc. 14th Int. Conf. on the Use of Computers in Radiation Therapy
Schreiner L J 2001 The potential for intensity modulated radiation therapy with cobalt-60
Insights 3 3
Schulz R J, Deye J A and Hendee W R 2001 Through the preoccupation with new technical
developments, physicists have lost sight of the realities of cancer care and statistics
Med. Phys. 28 21857
Schulz R J and Kagan A R 2002a (In regard to research in medical physics Int. J. Radiat.
Oncol. Biol. Phys. 49 89195) Int. J. Radiat. Oncol. Biol. Phys. 52 274
2002b On the role of intensity modulated radiation therapy in radiation oncology Med.
Phys. 29 147380
2003a Commentary on IMRT and cancer of the prostate Int. J. Radiat. Oncol. Biol.
Phys. 55 8512
2003b More precisely defined dose distributions are unlikely to affect cancer mortality
Med. Phys. 30 276
Schwarz M, Bos L, Lebesque J V, Mijnheer B J and Damen E M F 2002 The importance
of accurate dose calculations outside the beam edges in IMRT treatment planning
Schwarz M, Lebesque B J, Mijnheer B J and Damen E M F 2003 Influence of the volume
dependence parameter on treatment planning optimisation using the equivalent
uniform dose (EUD) Radiother. Oncol. 68 (Suppl. 1) S75
Schweikard A and Adler J R 2000 Motion tracking for respiratory motion in radiosurgery
Schweikard A, Glosser G, Bodduluri M, Murphy M and Adler J R 2000 Robotic motion
compensation for respiratory motion in radio-surgery Computed Aided Surgery 5
26377
Scielzo G, Stasi M, Chauvie S, Baiotto B, Malinverni G, Macias V, Munoz F and Gabriele
P 2002 IMRT sliding window with a 120-leaf MLC: dosimetry and first clinical
applications Radiother. Oncol. 64 (Suppl. 1) S236
Seco J, Evans P M and Webb S 2001 Analysis of the effects of the delivery technique on
an IMRT plan: comparison for multiple static field, dynamic and NOMOS MIMiC
collimation Phys. Med. Biol. 46 307387
444
References
Seco J, Verhaegen F and Bidmead M 2003 Development and commissioning of a Monte

Carlo dose calculation engine for clinical IMRT beams Proc. World Congress in
Medical Physics (Sydney) CDROM and website only
Seddon B, Bidmead M, Wilson J, Khoo V and Dearnaley D 2000 Target volume definition
in conformal radiotherapy for prostate cancer: quality assurance in the MRC RT-01
trial Radiother. Oncol. 56 7383
Seiler P G, Blattmann H, Kirsch S, Muench R K and Schilling Ch 2000 A novel tracking
technique for the continuous precise measurement of tumour positions in conformal
radiotherapy Phys. Med. Biol. 45 N10310
Sen A, Kelly D and Flynn J 2000 Verification of leaf patterns of a MIMiC collimator in
IMRT treatment Proc. World Congress of Medical Physics and the AAPM Annual
Seppenwoolde Y, Shirato H, Kitamura K, Shimizu S, van Herk M, Lebesque J V and
Miyasaka K 2001 3D Tumor motion in lung due to breathing and heartbeat, measured
during real-time tumor tracking radiation therapy Int. J. Radiat. Oncol. Biol. Phys. 51
(Suppl. 1) 24
2002 Precise and real-time measurement of 3D tumor motion in lung due to breathing
and heartbeat, measured during radiotherapy Int. J. Radiat. Oncol. Biol. Phys. 53
82234
Seppi E J, Munro P, Johnsen S W, Shapiro E G, Tognina C, Jones D, Pavkovich J M, Webb
C, Mollov I, Partain L D and Colbeth R E 2003 Megavoltage cone-beam computed
tomography using a high-efficiency image receptor Int. J. Radiat. Oncol. Biol. Phys.
55 793803
Serago C F, Chungbin S J, Buskirk S J, Ezzell G A, Collie A C and Vora S A 2002 Initial
experience with ultrasound localization for positioning prostate cancer patients for
external beam radiotherapy Int. J. Radiat. Oncol. Biol. Phys. 53 11308
Sethi A, Leybovich L, Dogan N and Emami B 2001a Matching tomographic IMRT fields
with static photon fields Med. Phys. 28 245965
Sethi A, Leybovich L B, Dogan N and Jeswani S 2001b The effect of heterogeneity
corrections on dose distributions when IMRT is used for lung cancer Radiother.
Sethi A, Leybovich L, Dogan N, Krasin M and Emami B 2000 Improvement of target
and critical structure doses in Corvus IMRT treatment plans Proc. World Congress of
Medical Physics and the AAPM Annual Congress, August 2000 paper TU-GBR-05
Sethi A, Mohideen N, Leybovich L, Dogan N and Mulhall J 2002 Role of IMRT in
reducing corporal bodies doses in dose escalation of prostate cancer Radiother.
Sethi A, Mohideen N, Leybovich L and Mulhall J 2003 Role of IMRT in reducing penile
doses in dose escalation for prostate cancer Int. J. Radiat. Oncol. Biol. Phys. 55 9708
Sham E, Hristov D and Fallone B 2001a Inverse planning by simulated annealing
minimization of dose-volume based objective function Med. Phys. 28 1206
Sham E, Hristov D, Stavrev P and Fallone G B 2001b Inverse treatment planning by
simulated annealing minimization of a dose-volume cost function Med. Phys. 28
1987
Sharp G, Jiang S B, Ruan D, Castanon and Shirato H 2003 Evaluation of prediction
methods for real-time tumour tracking during treatment Med. Phys. 30 1346
References
445
Sharpe M B 2003 Commissioning and quality assurance for IMRT treatment planning
IMRT- the state of the art: AAPM Medical Physics Monograph Number 29 ed
J R Palta and T R Mackie (Madison, WI: Medical Physics Publishing) pp 44973
Sharpe M B, Miller B M and Wong J W 2000a Compensation of x-ray beam penumbra in
conformal radiotherapy Med. Phys. 27 173945
Sharpe M B, Miller B M, Yan D and Wong J W 2000b Monitor unit settings for intensity
modulated beams delivered using a step-and-shoot approach Med. Phys. 27 271925
Shentall G S, Kirby M C, McKay D M, Harris F W and Read G 2001 Using biological
models as a way of determining the efficacy of IMRT for the prostate Radiother.
Shepard D, Earl M, Li X, Holmes T and Yu C 2001 Direct aperture optimization for step
and shoot IMRT Med. Phys. 28 1308
Shepard D M, Earl M A, Li X A, Naqvi S and Yu C 2002b Direct aperture optimisation: A
turnkey solution for step-and-shoot IMRT Med. Phys. 29 100718
Shepard D, Earl M, Naqvi S and Yu C 2002a An inverse planning tool for intensity
modulated arc therapy Med. Phys. 29 1336
Shepard D M, Earl M A, Yu C X and Xiao Y 2003a Aperture-based inverse planning
IMRTthe State of the Art: AAPM Medical Physics Monograph Number 29 ed
J R Palta and T R Mackie (Madison, WI: Medical Physics Publishing) pp 11537
Shepard D M, Jiang Z, Earl M A, Ferris M C, Lim J and Naqvi S 2003b A toolbox for
intensity modulated radiation therapy optimisation Med. Phys. 30 23202
Sherouse G W 2002 In regard to intensity-modulated radiotherapy collaborative working
group (Int. J. Radiat. Oncol. Biol. Phys. 2001 51 880914) Int. J. Radiat. Oncol. Biol.
Phys. 53 10889
Shih R, Li X A and Hsu W-L 2001 Dosimetric characteristics of dynamic wedged fields: a
Monte Carlo study Phys. Med. Biol. 46 N28192
Shimizu S, Shirato H, Kitamura K, Ogura S, Akita-Dosaka H, Tateishi U, Watanabe Y,
Fujita K Shimizu T and Miyasaka K 2000 Fluoroscopic real-time tumour-tracking
radiation treatment (RTRT) can reduce internal margin (IM) and set-up margin
(SM) of planning target volume (PTV) for lung tumours Proc. 42nd Annual ASTRO
Shimizu S, Shirato H, Ogura S, Akita-Dosaka H, Kitamura K, Nishioka T, Kagei K,
Nishimura M and Miyasaka K 2001 Detection of lung tumour movement in real-time
tumour-tracking radiotherapy Int. J. Radiat. Oncol. Biol. Phys. 51 30410
Shimoga K B, Kalend A M, Clark K and Greenberger J S 2001 SmartGate: A fuzzy-logic
based intelligent system for respiratory gating Int. J. Radiat. Oncol. Biol. Phys. 51
(Suppl. 1) 29
Shiomi H, Inoue T, Nakamura S and Inoui T 2001a Quality Assurance for an image-guided
frameless radiosurgery system using radiochromic film Radiation Medicine 18 107
13
Shiomi H, Inoue T, Nakamura S, Shimamoto S, Inoui T, Adler Jr. J R and Bodduluri M
2001b CyberKnife Japan J. Med. Phys. 21 1016 (in Japanese)
Shirato H, Seppenwoolde Y, Kitamura K, Onimura R and Shimizu S 2004 Intrafractional
tumor motion: lung and liver Seminars Radiat. Oncol. 14 1018
Shirato H, Shimizu S, Kitamura K, Nishioka T, Kagei K, Hashimoto S, Aoyama H,
Kuneida T, Shinohara N, Dosaka-Akita H and Miyasaka K 2000b Four-dimensional
treatment planning and fluoroscopic real-time tumor tracking radiotherapy for
moving tumour Int. J. Radiat. Oncol. Biol. Phys. 48 43542
446
References
Shirato H, Shimizu S, Kunieda T, Kitamura K, van Herk M, Kagei K, Nishioka T,

Hashimoto S, Fukota K, Aoyama H, Tsuchita K, Kudo K and Miyasaka K 2000a
Physical aspects of a real-time tumour-tracking system for gated radiotherapy Int. J.
Short R, Hoban P, Biggs D, Rose A and Smee B 2001 Pre-treatment dosimetric verification
of IMRT Radiother. Oncol. 61 (Suppl. 1) S108
Shou Z and Xing L 2004 Improve IMRT distribution by using spatially non-uniform
important factors Proc. 14th Int. Conf. on the Use of Computers in Radiation Therapy
Siebers J, Keall P, Kim J O, Arnfield M and Mohan R 2000 Dynamic IMRT Monte Carlo
dose calculation Proc. World Congress of Medical Physics and the AAPM Annual
Congress, August 2000 paper FR-EA309-05
Siebers J V, Lauterbach M, Keall P J and Mohan R 2002b Incorporating multi-leaf
collimator leaf sequencing into iterative IMRT optimisation Med. Phys. 29 9529
Siebers J, Lauterbach M, Tong S, Wu Q and Mohan R 2001a Reducing dose calculation
time for accurate iterative IMRT planning Med. Phys. 28 1308
2002a Reducing dose calculation time for accurate iterative IMRT planning Med.
Phys. 29 2317
Siebers J V and Mohan R 2003 Monte Carlo and IMRT IMRT- the state of the art: AAPM
Siebers J V, Tong S, Lauterbach M, Wu Q and Mohan R 2001b Acceleration of dose
calculations for intensity-modulated radiotherapy Med. Phys. 28 90310
Singh I, Ravindran B, Subramaniam B and John S 2002 Manual multi-leaf collimator for
60 Co beam shapinga feasibility study Med. Phys. 29 1213
Siochi R A C 1999 Minimising static intensity modulation delivery time using an intensity
solid paradigm. Int. J. Radiat. Oncol. Biol. Phys. 43 67180
2000a Simultaneous minimization of segments and junctions in virtual micro IMRT
2000 paper MO-EBR-04
2000b Getting started with IMRT RAD Magazine 26 (304) 478
2000c Virtual micro-intensity modulated radiation therapy Med. Phys. 27 248093
Sixel K E, Aznar M C and Ung Y C 2001 Deep inspiration breath hold to reduce irradiated
heart volume in breast cancer patients Int. J. Radiat. Oncol. Biol. Phys. 49 199204
Sohn J, Dempsey J, Drzymala R, Klein E, Grigereit T and Purdy J 2000b Analysis of
small beamlets for IMRT using radiochromic films Proc. World Congress of Medical
Physics and the AAPM Annual Congress, August 2000 paper MO-GBR-02
Sohn J, Dempsey J, Low D, Klein E, Dryzmala R, Kong F and Mansur D 2001a IMRT
to breast tissue and associated lymph nodes with consideration of breathing motion
Med. Phys. 28 1253
Sohn J W, Dempsey J F, Suh T S and Low D A 2003 Analysis of various beamlet sizes for
IMRT with 6 MV photons Med. Phys. 30 24329
Sohn J W, Low D A and Deasy J O 2001b Motion-induced distribution artifacts in lung
cancer IMRT Int. J. Radiat. Oncol. Biol. Phys. 51 3478
Sohn J W, Low D A and Klein E E 2000a Dynamic multileaf collimator performance
for intensity modulated radiation therapy Proc. 42nd Annual ASTRO Meeting. Int. J.
Radiat. Oncol. Biol. Phys. 48 (Suppl.) 341
References
447
Solberg T, Agazaryan N, Tenn S, Wink N, DeMarco J, Medin P, Lee S and Parker R

2003 Initial experience with intensity modulated radiotherapy for prostate cancer
delivered using a micro-multileaf collimator. Proc. World Congress in Medical
Physics (Sydney) CDROM and website only
Solberg T, Paul T, Boone R, Agazaryan N, Urmanita T, Arellano A, Llacer J, Fogg R,
DeMarco J and Smathers J 2000 Feasibility of gated IMRT Proc. World Congress of
Medical Physics and the AAPM Annual Congress, August 2000 paper TH-EBR-02
Sonke J-J, Ploeger L S, Brand B, Smitsmans M H P and van Herk M 2003 Geometrical
verification of dynamic intensity modulated radiotherapy using an amorphous silicon
flat panel imager Med. Phys. 30 1352
2004 Leaf trajectory verification during dynamic intensity modulated radiotherapy
using an amorphous silicon flat panel imager Med. Phys. 31 38995
Sorvari P, Hamalainen T, Sievinen J, Pyykkonen J and Tenhunen M 2000 Commissioning
of the dynamic multileaf collimator system for clinical use in a 6 MV linear
accelerator Radiother. Oncol. 56 (Suppl. 1) S194
Soubra M, Stablein C, Dalope M G, Elbadawi N and El-Ghandour O 2001a Prostate
immobization and an alignment system for IMRT patients Radiother. Oncol. 61
(Suppl. 1) S109
2001b IMRT prostate patient immobilization and alignment system Med. Phys. 28
1209
Spirou S V and Chui C S 1994 Generation of arbitrary intensity profiles by dynamic jaws
or multileaf collimators Med. Phys. 21 103141
1996 Generation of arbitrary intensity profiles by combining the scanning beam with
dynamic multileaf collimation Med. Phys. 23 18
1998 A gradient inverse-planning algorithm with dose-volume constraints Med. Phys.
25 32133
Spirou S V, Fournier-Bidoz N, Yang J, Chui C-S and Ling C C 2001 Smoothing intensitymodulated beam profiles to improve the efficiency of delivery. Med. Phys. 28 2105
12
Spirou S, Yang J, Chui C and Ling C, 2000 Smoothing intensity modulated beam profiles
to facilitate delivery Proc. World Congress of Medical Physics and the AAPM Annual
Congress, August 2000 paper MO-EBR-02
Spitters-Post I, van der Vight L P, van Lin E N J T and Visser A G 2002 The use of
fiducial (gold) markers for monitoring prostate movements during radiation therapy
Staffurth J, Dearnaley D, Convery D, Adams E, Clark C, Huddart R and Webb S 2003a
A planning study to compare the plan quality and delivery efficiency of three IMRT
planning and delivery system Clinical Oncol. 15 (Suppl. 2) S334
Staffurth J, Dearnaley D, McNair H, Convery D, Adams E, Clark C, Bidmead M,
Warrington A P, Huddart R, Webb S and Horwich A 2003b Early results of a Phase
2 trial of pelvic nodal irradiation in prostate cancer IMRT Clinical Oncol. 15 (Suppl.
2) S17
Staffurth J, Dearnaley D, McNair H, Convery D, Adams E, Clark C, Vaarkamp J,
Huddart R, Webb S and Horwich A 2002 Early results of a phase 1 trial of pelvic
nodal irradiation in prostate cancer with intensity modulated radiotherapy (IMRT)
Starkshall G, Sherouse G W and Hendee W R 2001 The future will not need clinical
therapy physicists Med. Phys. 28 8657
448
References
Stasi M, Baiotto B, Malinverni G, Gabriele P and Scielzo G 2000 Multiple static field
technique in prostate cancer: a pilot study using a traditional 3D treatment planning
system Radiother. Oncol. 56 (Suppl. 1) S110
Stavrev P V, Hristov D H and Sham E 2001 IMRT inverse treatment planning optimisation
based on physical constraints and biological objectives Med. Phys. 28 1987
Steenbakkers R J H M, Deurloo K E I, Nowak P J C M, van Herk M and Rasch C R N
2002 Comparison of CT and MRI delineation for 3-D conformal treatment planning
of the prostate Radiother. Oncol. 64 (Suppl. 1) S274
Stein J, Bortfeld T, Dorschel B and Schlegel W 1994 Dynamic x-ray compensation for
conformal radiotherapy by means of multileaf collimation Radiother. Oncol. 32 163
73
Stein J, Mohan R, Wang X-H, Bortfeld T, Wu Q, Preiser K, Ling C C and Schlegel W 1997
Number and orientation of beams in intensity-modulated radiation treatments Med.
Phys. 24 14960
Steinberg T, Allmendinger K, Dudas A and Calderon E 2002 Dose stability for low monitor
unit treatments Med. Phys. 29 1212
Sternick E S, Bleier A R, Carol M P, Curran B H, Holmes T W, Kania A A, Lalonde R and
Larson L S 1997 Intensity modulated radiation therapy: what photon energy is best?
Proc 12th International Conference on the Use of Computers in Radiation Therapy,
Salt Lake City, Utah, May 1997 ed D D Leavitt and G Starkschall (Madison, WI:
Stevens C W 2003 Lung cancer radiotherapy IMRTthe State of the Art: AAPM Medical
Strom E A 2002 Breast IMRT: New tools leading to new vision Int. J. Radiat. Oncol. Biol.
Phys. 54 12978
Stromberg J S, Sharpe M B, Kim L H, Kini V J, Jaffray D A, Martinez A A and Wong J
W 2000 Active breathing control (ABC) for Hodgkins disease: reduction in normal
tissue irradiation with deep inspiration and implications for treatment Int. J. Radiat.
Subramanian T R, Gibbons J P and Hendee W R 2002 Linear accelerators used for IMRT
should be designed as small field high intensity intermediate energy units Med. Phys.
29 25268
Suh Y, Yi B, Ahn S, Kim J, Lee S, Shin S and Choi E 2004 Aperture manoeuver with
compelled breath (AMC) for moving tumors: a feasibility study using moving
phantom Med. Phys. 31 7606
Suh Y, Yi B, Ahn S, Lee S, Kim J, Shin S and Choi E 2003 Adaptive field shaping to
moving tumor with the compelled breath control: a feasibility study using moving
phantom Proc. World Congress in Medical Physics (Sydney) CDROM and website
only
Suh Y, Yi B, Kim J, Ahn S, Lee S, Shin S and Choi E 2002 Prediction of the tumour
location in the lung from the skin motion Med. Phys. 29 1240
Suit H 2002 The Gray Lecture 2001: coming technical advances in radiation oncology Int.
Sun X and Xia P 2004 A new smoothing procedure to reduce delivery segments for static
MLC-based IMRT planning Med. Phys. 31 115865
References
449
Svatos M, Rosenman J, Cullip T, Verhey L and Hughes J 2000 Mixing electrons with
intensity modulated photon beams to reduce integral dose Proc. World Congress of
Medical Physics and the AAPM Annual Congress, August 2000 paper WE-E313-10
Svensson R, Kallman P and Brahme A 1994 Analytical solution for the dynamic control
of multileaf collimators Phys. Med. Biol. 39 3761
Swerdloff S, Mackie T R and Holmes T 1994a Method and apparatus for radiation therapy
US Patent 5,317,616
1994b Multi-leaf radiation attenuator for radiation therapy US Patent 5,351,280
Swindell W, Simpson R G and Oleson J R 1983 Computed tomography with a linear
accelerator with radiotherapy applications Med. Phys. 10 41620
Swinnen A, van Esch A and Huyskens D P 2002 Comparison of static and dynamic
delivery of intensity-modulated radiation therapy Radiother. Oncol. 64 (Suppl. 1)
S107
Szanto J, Malone S, Gerig L, Punkari and Hube P 2002 A system for image-guided prostate
Tanner S F, Finnigan D J, Khoo V S, Mayles P, Dearnaley D P, and Leach M O 2000
Radiotherapy planning of the pelvis using distortion corrected MR images: the
removal of system distortion Phys. Med. Biol. 45 211732
Tao L, Rakfal S and Wu A 2002 Comparison of integral doses in conventional 2D,
conformal 3D and IMRT plans Med. Phys. 29 1211
Teh B S, Bastasch M D, Wheeler T M, Mai W-Y, Frolov A, Uhl M, Lu H H, Carpenter L S,
Chiu J K, McGary J, Woo S Y, Grant W H III and Butler E B 2003 IMRT for prostate
cancer: defining target volume based on correlated pathologic volume of disease Int.
Teh B S, Mai W Y, Huang E, Carpenter L S, Lu H H, Chiu J K, Woo S Y, Grant W
H and Butler E B 2001b Late gastrointestinal (GI) and genitourinary (GU) toxicity
following intensity modulated radiation therapy (IMRT) for prostate cancer Int. J.
Teh B S, Mai W-Y, Uhl B M, Augspurger M E, Grant W H III, Lu H H, Woo S Y, Carpenter
S L, Chiu J K and Butler E B 2001a Intensity-modulated radiation therapy (IMRT)
for prostate cancer with the use of a rectal balloon for prostate immobilization: acute
toxicity and dose-volume analysis Int. J. Radiat. Oncol. Biol. Phys. 49 70512
Tepper J 2000 Form and function: the integration of physics and biology Int. J. Radiat.
Teslow T N 2002 A software program that evaluates treatment plans for intensity
modulated radiotherapy Radiother. Oncol. 64 (Suppl. 1) S51
Thieke C, Bortfeld T and Kuefer K-H 2003a New optimization concepts in inverse
Thieke C, Bortfeld T, Niemierko A, Kuefer K-H and Nill S 2003b Multicriteria
optimisation in inverse radiotherapy planning Radiother. Oncol. 68 (Suppl. 1) S44
Thieke C, Bortfeld T, Niemierko A and Nill S 2003c From physical dose constraints to
equivalent uniform dose constraints in inverse radiotherapy planning Med. Phys. 9
23329
Thieke C, Nill S, Oelfke U and Bortfeld T 2001 Acceleration of IMRT dose calculations
by importance sampling of calculation matrices Radiother. Oncol. 61 (Suppl. 1) S46
2002 Acceleration of intensity-modulated radiotherapy dose calculation by
importance sampling of the calculation matrices Med. Phys. 29 67681
450
References
Thilmann C, Krempien R, Hoess A, Sroka-Perez G, Schramm O, Wannenmacher M and

Debus J 2002 Inversely planned intensity modulated radiotherapy (IMRT) of the
breast and the locoregional lymph nodes Radiother. Oncol. 64 (Suppl. 1) S56
Thomas M D R and Symonds-Tayler J R N 2003 Characterization of a transmission
ionisation beam-imager for radiotherapy verification Phys. Med. Biol. 48 263344
Ting J Y, Butker E, Ghavidel S and Davis L W 2000a A stop and shoot leaf-sequencing
algorithm for IMRT Proc. World Congress of Medical Physics and the AAPM Annual
Ting J, Iwinski A, Papanikolaou N, Ghavidel S and Davis L 2000b A comparison between
step-and-shoot and sliding windows leaf sequencing algorithms in IMRT Proc.
Tome W A, Meeks S L, McNutt T R, Buatti J M, Bova F J, Friedman W A and Mehta M
2001 Optically guided intensity modulated radiotherapy Radiother. Oncol. 61 3344
Tome W A, Mehta M P, Meeks S L, Buatti J M, Bova F J and Friedman W A 2000 Image
guided fractionated stereotactic radiotherapy Radiother. Oncol. 56 (Suppl. 1) S60
Toner S, Shih J, Boccuzzi D, Chiu-Tsao S, Gliedman P and Harrison L 2000 The use of
external MR- and CT-compatible fiducial markers for image fusion in stereotactic
Topolnjak R, van der Heide U A and Lagendijk J J W 2004a IMRT sequencing for sixbank multi-leaf collimator system Proc. 14th Int. Conf. on the Use of Computers in
Topolnjak R, van der Heide U A, Raaymakers B W, Kotte A N T J and Lagendijk J J
W 2003b IMRT sequencing for a six-bank multi-leaf system Radiother. Oncol. 68
(Suppl. 1) S11415
Topolnjak R, van der Heide U A, Raaymakers B, Kotte A, Welleweerd J and Lagendijk
J J W 2002 Six banks multi-leaf system, high resolution and large field size MLC
2003a Six-bank multi-leaf system for IMRT Phys. Medica. 19 59
2004b A six-bank multi-leaf system, high precision shaping of large fields Phys. Med.
Biol. 49 264556
Trapp J V 2003 Imaging and radiation interactions of polymer gel dosimeters PhD Thesis
University of Queensland, Australia
Trapp J V, Michael G, Evans P M, Baldock C, Leach M O and Webb S 2004b Dose
resolution in gel dosimetry: effect of uncertainty in the calibration function Phys.
Med. Biol. 49 N13946
Trapp J, Partridge M, Hansen V, Childs P, Bedford J, Warrington J, Leach M O and Webb
S 2004a The use of gel dosimetry for verification of electron and photon treatment
plans in carcinoma of the scalp Phys. Med. Biol. 49 162535
Trichter F and Ennis R D 2001 Patient positioning in prostate IMRT: Advantages of BAT
over EPID Med. Phys. 28 1292
Tsai J-S, Rivard M J and Engler M J 2000 Dependence of linac output on the switch rate
of an intensity-modulated tomotherapy collimator Med. Phys. 27 221535
Tsunashima Y, Sakae T, Shioyama Y, Kagei K, Terunuma T and Akine Y 2003 Correlation
between respiratory wave form using a respiratory sensor and the 3D tumor motion
in respiratory gated radiotherapy Proc. World Congress in Medical Physics (Sydney)
Tucking T, Nill S, Thilmann C and Oelfke U 2003 Application of a new external MMLC
for high precision IMRTa feasibility study Radiother. Oncol. 68 (Suppl. 1) S75
References
451
Turian J and Smith B 2002 Evaluation of dose differences between multiple static segment
(MSS) and sliding window (SW) IMRT techniques Med. Phys. 29 1261
Twyman N and Thomas S 2002 HD-270turning 1 cm multileaf collimators into 2 mm
multileaf collimators RAD Magazine 28 (330) 1415
Unkelbach J and Oelfke U 2003 Inclusion of stochastic organ movements in IMRT
2004 Organ movements in IMRT treatment planning: inverse planning based on
probability distributions Proc. 14th Int. Conf. on the Use of Computers in Radiation
Ung Y C, Caldwell C B, Mah K, Danjoux C E, Balogh J B, Ganguli S N, Tirona R and
Ehrlich L E 2000 Fusing 18 flourodeoxyglucose (FDG)-hybrid PET to CT images
significantly alters treatment planning in the radical treatment of non-small cell
carcinoma of the lung Proc. 42nd Annual ASTRO Meeting. Int. J. Radiat. Oncol.
Uy N W, Woo S Y, Teh B S, Mai W-Y, Carpenter L S, Chiu J K, Lu H H, Gildenberg
P, Trask T, Grant W H and Butler E B 2002 Intensity-modulated radiation therapy
(IMRT) for meningioma Int. J. Radiat. Oncol. Biol. Phys. 53 126570
Vaarkamp J, Warrington A P, and Dearnaley D P 2003 A forward planned conformal
class solution to generate concave dose distribution to treat prostate and pelvic nodes
Valinta D, Dejean C, Erenaud S and Floiras J L 2001a Quality control of the IMRT
Valinta D, Dejean C and Floiras JL 2001b Inverse planning and IMRT delivery quality
control Phys. Medica. 17 114
Valinta D, Migdal S, de Carlan L and Floiras J L 2000 First clinical application of IMRT
using an inverse plan in St. Cloud Radiother. Oncol. 56 (Suppl. 1) S196
Van Asselen B, Dehnad H, Raaijmakers C P J, Terhaard C H J and Lagendijk J J W
2001a Intensity modulated radiotherapy for oropharyngeal cancer: a class solution
Van Asselen B, Dehnad H, Terhaard C H J, Lagendijk J J W and Raaijmakers C P J, 2004
Segmental IMRT for oropharyngeal cancer in a clinical setting Radiother. Oncol. 69
25966
Van Asselen B, Raaijmakers C P J, Hofman P and Lagendijk J J W 2001b An improved
breast irradiation technique using three-dimensional geometrical information and
intensity modulation Radiother. Oncol. 58 3417
Van Dalen S, Keijzer M, Huizenga H and Storchi P R M 2000 Optimization of multileaf
collimator settings for radiotherapy treatment planning Phys. Med. Biol. 45 361525
Van den Heuvel F, Powell T, Seppi E, Littrupp P, Khan M, Wang Y and Forman JD 2003
Independent verification of ultrasound based image-guided radiation treatment, using
electronic portal imaging and implanted gold markers Med. Phys. 30 287887
Van der Heide U A, Hoekstra A, Kotte A, van Rijk P P, Lagendijk J J W, Terhaard C H J
2003 The use of PET imaging in radiotherapy dose painting Phys. Medica. 19 44
Van der Heide U, Nederveen A, Hofman P, Welleweerd J and Lagendijk J 2001 The
microboost approach to IMRT of the prostate Med. Phys. 28 1261
Van Dieren E B, Nowak P J C M, Wijers O B, van Sornsen de Koste J R, van der Est H,
Binnekamp D P, Heijmen B J M and Levendag P C 2000 Beam intensity modulation
using tissue compensators or dynamic multileaf collimation in three-dimensional
452
References
conformal radiotherapy of primary cancers of the oropharynx and larynx, including

the elective neck Int. J. Radiat. Oncol. Biol. Phys. 47 1299309
Van Dyk J, Kron T, Grigorov G, Chen J, Wong E, Yu E, Lock M, Rodrigues G and Bauman
G 2003 Early experience with helical tomotherapy Radiother. Oncol. 68 (Suppl. 1)
S39
Van Esch A, Sommer P D T, Verstraete J and Huyskens D 2001b QA of IMRT and DMLC
by means of the Varian PortalVision system: the clinical implementation. Radiother.
Van Esch A, Vanstraelen B, Verstraete J, Kutcher G and Huyskens D 2001a Pre-treatment
dosimetric verification by means of a liquid-filled electronic portal imaging device
during dynamic delivery of intensity modulated treatment fields Radiother. Oncol. 60
18190
Van Herk M 2000 The role of multimodality imaging in radiotherapy Radiother. Oncol. 56
(Suppl. 1) S16
2003 Applications of image registration in radiotherapy Phys. Medica. 19 445
Van Herk M, Artignan X, de Bola J, Smitsmans M H P, Lebesque J V and Jaffray D A
2002b Automatic prostate localization for image-guided radiotherapy of the prostate
Van Herk M, Betgen A, Brand B, Jaffray D, Remeijer P, Smitsmans M, Sonke J-J and Zijp
L 2004 An integrated system for on-line cone-beam CT image-guided radiotherapy
software aspects Proc. 14th Int. Conf. on the Use of Computers in Radiation Therapy
Van Herk M, de Bois J, de Munck J, de Jaeger K, Lebesque J and Rasch C 2003 Use
of multi-modality imaging to improve target volume delineation accuracy Clinical
Van Herk M, Remeijer P, Rasch C and Lebesque J V 2000 The probability of correct target
dosage: dose-population histograms for deriving treatment margins in radiotherapy
Van Herk M, Scheider C, Sonke J J, Damen E and DeJaeger K 2002a Respirationcorrelated CT of lung cancer patients Radiother. Oncol. 64 (Suppl. 1) S801
Van Nimwegen A, Lagerwaard F, de Pan C, Nowak P and Levedag P 2001 IMRT for the
nasopharynx, a planning study Radiother. Oncol. 61 (Suppl. 1) S47
Vanregemorter J, de Ost B and van den Weyngaert D 2000 Forward IMRT planning: a
phantom study Radiother. Oncol. 56 (Suppl. 1) S97
Vanregemorter J, Van Gestel D and van den Weyngaert D 2002 Introduction of multimodality imaging for conformal prostate radiation treatment Radiother. Oncol. 64
(Suppl. 1) S268
Van Santvoort J P C and Heijmen B J M 1996 Dynamic multileaf collimation without
tongue-and-groove underdosage effects Phys. Med. Biol. 41 2091105
Van Sornsen de Koste J R, Lagerwaard F J, de Boer H C J, Nijssen-Visser M R J and Senan
S 2003a Are multiple CT scans required for planning curative radiotherapy in lung
tumours of the lower lobe? Int. J. Radiat. Oncol. Biol. Phys. 55 13949
Van Sornsen de Koste, Lagerwaard F J, Nijssen-Visser M R J, Graveland W J and Senan S
2003b Tumor location cannot predict the mobility of lung tumours: a 3D analysis of
data generated from multiple CT scans Int. J. Radiat. Oncol. Biol. Phys. 56 34854
Van Vliet-Vroegindeweij C, Louwe R, Dwarswaard M, van Ingen K, Schneider C, Minken
A, van Herk M and Mijnheer B 2001 Dosimetric verification of IMRT using an
electronic portal imaging device Radiother. Oncol. 61 (Suppl. 1) S107
References
453
Vedam S S, Keall P J, Kini V R and Mohan R 2001 Determining parameters for respirationgated radiotherapy Med. Phys. 28 213946
Vedam S, Keall P, Kini V, Mostafavi H and Mohan R 2003b Predicting breathing motion
for 4D radiotherapy (WIP) Med. Phys. 30 1472
Vedam S S, Kini V R, Keall P J, Ramakrishnan V, Mostafavi H and Mohan R 2003a
Quantifying the predictability of diaphragm motion during respiration with a noninvasive external marker Med. Phys. 30 50513
Verellen D 2001 Delivery techniques of IMRT: tomotherapy and miniMLCs Radiother.
Verellen D, Linthout N, Soete G, van Acker S, de Roover P and Storme G 2002
Considerations on treatment efficiency of different conformal radiation therapy
techniques for prostate cancer Radiother. Oncol. 63 2736
Verellen D, Linthout N and Storme G 2000 Comparison between different approaches for
conformal treatment of the prostate: A theoretical and dosimetric evaluation Proc.
paper FR-EABR-05
Verellen D, Soete G, Linthout N, Van Acker S and Storme G 2001 Improved target
localization and set-up for CFRT/IMRT of the prostate by combining real-time
infrared tracking and stereoscopic X-ray imaging Med. Phys. 28 1295
Vergote K, de Deene Y, Duthoy W, de Gersem W, de Neve W, Achten E and de Wagter C
2004 Validation and application of polymer gel dosimetry for the dose verification of
an intensity modulated arc therapy (IMAT) treatment Phys. Med. Biol. 49 287305
Verhaegen F and Liu H H 2001 Incorporating dynamic collimator motion in Monte Carlo
simulations: an application in modelling a dynamic wedge Phys. Med. Biol. 46 287
96
Verhey L 2000 IMRT: applications and benefits in comparison with 3D conformal
2001 IMRT with multileaf collimators Med. Phys. 28 1241
Vervoort C, Verstraete J, Vanstraelen B, Vandevelde G, Van Esch A, Scherpenberg J and
Huyskens D 2001 Implementation of a patient set-up verification and correction
protocol with an on-line portal imaging in a IMRT procedure for localised prostate
cancer Radiother. Oncol. 61 (Suppl. 1) S47
Vial P, Bromley R and Oliver L 2003 Commissioning of a dynamic multileaf collimator
delivery system Proc. World Congress in Medical Physics (Sydney) CDROM and
website only
Vicini F A, Sharpe M, Kestin L, Martinez A, Mitchell C K, Wallace M F, Matter R and
Wong J 2002 Optimising breast cancer treatment efficacy with intensity modulated
radiotherapy Int. J. Radiat. Oncol. Biol. Phys. 54 133644
Vicini F A, Sharpe M, Kestin L, Wong J, Remouchamps V and Martinez A 2004 The use
of intensity modulated radiation therapy in the treatment of breast cancer: evolving
definition, misdirected criticism, and untoward effects Int. J. Radiat. Oncol. Biol.
Phys. 58 16424
Vieira S C, Dirkx M L P, Pasma K L and Heijmen B J M 2001 Fast and accurate verification
of leaf calibration for dynamic multileaf collimation with an EPID Radiother. Oncol.
61 (Suppl. 1) S17
Vineberg K A, Eisbruch A, Coselmon M M, McShan D L, Kessler M L and Fraass B A
2002 Is uniform target dose possible in IMRT plans in the head and neck? Int. J.
454
References
Vineberg K A, Eisbruch A, Kessler M L, McShan D L, Martel M K and Fraass B F 2000

Parotid sparing without sacrificing target dose uniformity using optimized beamlet
IMRT Proc. 42nd Annual ASTRO Meeting. Int. J. Radiat. Oncol. Biol. Phys. 48
(Suppl.) 174
Vineberg K A, McShan D L, Kessler M L and Fraass B A 2001 Comparison of dose,
dose-volume and biologically-based cost functions for IMRT plan optimization Int.
J. Radiat. Oncol. Biol. Phys. 51 (Suppl. 1) 712
Visser A G, Stroom J C and Huizenga H 2000 Target volume margins based on measured
uncertainties Radiother. Oncol. 56 (Suppl. 1) S54
Visser A G, van Lin E N J T, Welmers A, Futterer J, Barentsz J O, van der Vight L P
and van der Meazen R W M 2002 Fiducial markers for MRI-based target volume
definition and treatment verification Radiother. Oncol. 64 (Suppl. 1) S41
Vuong T, Belliveau P, Falco T, Meterissian S and Wexler M 2002 Clinical application
of IMRT in the treatment of patients with unresectable T4 or recurrent rectal or
rectosigmoid cancer: feasibility and toxicity data Radiother. Oncol. 64 (Suppl. 1)
S56
Wagner T H, Meeks S L, Bova F J, Friedman W A, Buatti J M and Bouchet L G 2001
Isotropic beam bouquets for shaped linear accelerator radiosurgery Phys. Med. Biol.
46 257186
Wagman R, Yorke E, Giraud P, Ford E, Sidhu K, Mageras G, Minsky B and Rosenzweig
K 2001 Reproducibility of organ position with respiratory gating for liver tumours:
Use in dose-escalation Int. J. Radiat. Oncol. Biol. Phys. 51 (Suppl.) 28
Wahab S, Low D, El Naqa I, Parikh P, Nystrom M, Bradley J, Mutic S and Dempsey J
2003 Use of four-dimensional computed tomography in conformal therapy planning
for lung cancer Med. Phys. 30 1364
Wahab S H, Malyapa R S, Mutic S, Grigsby P W, Deasy J O, Miller T R, Zoberi I
and Low D A 2004a Intensity modulated radiation therapy for cervical cancer: a
treatment planning comparison against brachytherapy Proc. 14th Int. Conf. on the
Use of Computers in Radiation Therapy (Seoul, May) p 6925
Wahab S H, Parikh P, Nystrom M, El Naqa I, Hubenschmidt J, Pierburg B, Lu W, Bradley
J D and Low D A 2004b Treatment planning using four dimensional computed
tomography Proc. 14th Int. Conf. on the Use of Computers in Radiation Therapy
(Seoul, May) pp 626
Walker V 2001 Conformal radiotherapy treatment delivery RAD Magazine 27 (233) 256
Wang C, Dai J and Hu Y 2003a Optimization of beam orientations and beam weights
for conformal radiotherapy using mixed integer programming Phys. Med. Biol. 48
406576
Wang D, Hill R W and Lam S 2004 A new algorithm for determining collimator angles
that favour efficiency in MLC based IMRT delivery Med. Phys. 31 124953
Wang J, Sabbas A, Kulidzhanov F, Presser J, Vallejo A and Nori D 2002b Comparison
study of IMRT treatment with 3D-RTP in treating ethmoid cancer Med. Phys. 29
1260
Wang L, Jacob R, Movsas B, Chen L, Fourkal E, Ma C and Pollack A 2003b Dosimetric
comparison of different multileaf collimator widths in the treatment of prostate
cancer with intensity modulated radiotherapy Med. Phys. 30 1405
Wang L, Jacob R, Movsas B, Fourkal E, Chen L, Konoski A, Feigenberg S P A and Ma C
2003c Dosimetric advantage of using small-width multi-leaf collimator in the IMRT
of prostate cancer Radiother. Oncol. 68 (Suppl. 1) S99
References
455
Wang L, Movsas B, Jacob R and Ma C 2003d Dosimetric advantage and clinical

implication of using a micro multileaf collimator in the treatment of prostate cancer
with intensity modulated radiotherapy Proc. World Congress in Medical Physics
(Sydney) CDROM and website only
Wang L, Yorke E and Chui C-S 2002a Monte Carlo evaluation of 6 MV intensity modulated
radiotherapy plans for head and neck and lung treatments Med. Phys. 29 270517
Warkentin B, Steciw S, Rathee S and Fallone B G 2003 Dosimetric IMRT verification with
a flat-panel EPID Med. Phys. 30 314355
Warrington A P and Adams E J 2001 Conformal and intensity modulated radiotherapy
using cobalt-60 and 6 MV X-ray beams: a treatment planning comparison of different
sites Radiother. Oncol. 61 (Suppl. 1) S73
2002 Cobalt-60 teletherapy for cancer: a revived treatment modality for the
21st century Proc. Seminar on Appropriate Medical Technology for Developing
Countries; IEE Feb 6th pp 21.121.19
Warrington A P, Robinson M, Smith N, Adams E J, Hansen V N, and Partridge M
2002 Conformal and intensity modulated radiotherapy on a cobalt unit Proc. IPEM
Biennial Meeting (Southampton, 1617 July) p 37
Wavelength 2000a IMRT update from ASTRO99 Wavelength 4 (1) 14
2000b Phased field technique enables virtual 160 leaf MLC for finer field shaping
Wavelength 4 (1) 1217
2000c Intensity modulated arc therapy looks promising in worlds first clinical trial
Wavelength 4 (2) 15
2000d IMRT international symposium provides state-of-the technique update
2001a New technology promises to redefine precision in radiotherapy Wavelength 5
(2) 714
2001b Ghent university hospitals class solution approach streamlines IMRT for
ethmoid sinus cancer Wavelength 5 (3) 14
2001c Collaboration results in new IMRT solution Wavelength 5 (3) 910
2001d University medical centre Utrecht begins IMRT protocol for prostate cancer
Wavelength 5 (3) 56
2001e Elekta users to demonstrate IMRT employing commercial solutions
2002a Active breathing coordinator brings greater precision to radiotherapy
2002b Hospitals integrating Active Breathing Coordinator into clinical protocols
Wavelength 6 (2) 17
2002c Christie hospital among first to evaluate new Elekta technology Wavelength 6
(3) 14
2002d Active breathing coordinator freezes targets in extracranial stereotactic
radioablation Wavelength 6 (3) 1013
2003a Beaumont work results in breakthrough Elekta Synergy system Wavelength 7
(1) 15
2003b Clatterbridge Centre for Oncology pursues goal of better prostate cancer
treatment with IMRT Wavelength 7 (2) 18
2003c PrecisePLANmaking the complex simple. Aperture-based inverse planning
reduces MUs, segments Wavelength 7 (2) 14
456
References
2003d Clinical feasibility studies continue to show value of active breathing

coordinator Wavelength 7 (2) 1418
2003e Wavelength 7 (3) 116
2004 Wavelength 8 (1) 119
Webb S 1989 Optimisation of conformal radiotherapy dose distributions by simulated
annealing Phys. Med. Biol. 34 134969
1991a Optimisation by simulated annealing of three-dimensional conformal treatment
planning for radiation fields defined by a multileaf collimator Phys. Med. Biol. 36
120126
1991b Optimisation of conformal radiotherapy dose distributions by simulated
annealing 2: inclusion of scatter in the 2D technique Phys. Med. Biol. 36 122737
1992 Optimisation by simulated annealing of three-dimensional conformal treatment
planning for radiation fields defined by a multileaf collimator: 2. Inclusion of twodimensional modulation of the x-ray intensity Phys. Med. Biol. 37 1689704
1993 The Physics of Three-Dimensional Radiation Therapy: Conformal Radiotherapy, Radiosurgery and Treatment Planning (Bristol: IOP Publishing)
1995 The problem of isotropically orienting N converging vectors in space with
applications to radiotherapy planning Phys. Med. Biol. 40 94554
1997 The Physics of Conformal Radiotherapy: Advances in Technology (Bristol: IOP
Publishing)
1998a Configuration options for intensity-modulated radiation therapy using
multiple-static fields shaped by a multileaf collimator Phys. Med. Biol. 43 24160
1998b Configuration options for intensity-modulated radiation therapy using
multiple-static fields shaped by a multileaf collimator. 2constraints and limitations
on 2D modulation Phys. Med. Biol. 43 148195
1999 Conformal intensity-modulated radiotherapy (IMRT) delivered by robotic
linactesting IMRT to the limit? Phys. Med. Biol. 44 163954
2000a Optimisation of IMRT: can the ultimate be established and what are the
consequences of approximation? Radiother. Oncol. 56 (Suppl. 1) S96
2000b Regarding x-ray leakage during IMRT Brit. J. Radiol. 73 1339
2000c A new concept of multileaf collimator (the shuttling MLC)an interpreter for
high-efficiency IMRT Phys. Med. Biol. 45 334358
2000d Intensity Modulated Radiation Therapy (Bristol: IOP Publishing)
2000e Conformal intensity-modulated radiotherapy (IMRT) delivered by robotic
linacconformality versus efficiency of dose delivery Phys. Med. Biol. 45 171530
2001a Concepts for shuttling multileaf collimators for intensity-modulated radiation
therapy Phys. Med. Biol. 46 63751
2001b A simple method to control aspects of fluence modulation in IMRT planning
Phys. Med. Biol. 46 N18795
2001c The philosophy of IMRT Radiother. Oncol. 61 (Suppl. 1) S12
2002a Intensity-modulated radiation therapy using only jaws and a mask Phys. Med.
Biol. 47 25775
2002b Intensity-modulated radiation therapy using only jaws and a mask 2: a
simplified concept of single bixel attenuators Phys. Med. Biol. 47 186979
2002c Fewer segments for IMRT generated by modulation splitting Phys. Med. Biol.
47 N21722
2002d Some snapshots from the history of radiotherapy physics SCOPE 11 812
References
457
2002e The future of external beam radiotherapy: the dream and the reality Phys.
Medica. 17 20715
2003a The history of IMRT IMRTthe State of the Art: AAPM Medical Physics
Publishing) pp 123
2003b The physical basis of IMRT and inverse planning Brit. J. Radiol. 76 67889
2004a A gravity-oriented-device for IMRT can never rival other IMRT delivery
methods Phys. Med. Biol. 49 L59
2004b Direct aperture optimisation for a variable aperture collimator for intensity
modulated radiation therapy Phys. Med. Biol. 49 N4755
Webb S, Bortfeld T, Stein J and Convery D 1997 The effect of stair-step leaf transmission
on the tongue-and-groove problem in dynamic radiotherapy with a multileaf
collimator Phys. Med. Biol. 42 595602
Webb S, Convery D J and Evans P M 1998 Inverse planning with constraints to generate
smoothed intensity modulated beams Phys. Med. Biol. 43 278594
Webb S, Cosgrove V and Evans P M 2001 Royal Marsden Hospital in Phase 1 IMRT
clinical trial for prostate cancer Wavelength 5 (1) 15
Webb S, Hartmann G, Eschner G and Schlegel W 2003 Intensity modulated radiation
therapy using a variable aperture collimator Phys. Med. Biol. 48 122338
Webb S and Nahum A E 1993 The biological effect of inhomogeneous tumour irradiation
with inhomogeneous clonogenic cell density Phys. Med. Biol. 38 65366
Webb S and Oldham M 1996 A method to study the characteristics of 3D dose distributions
created by superposition of many intensity-modulated beams via a slit aperture with
multiple absorbing vanes Phys. Med. Biol. 41 213553
West M and Jones L 2000 Using MLCs to deliver compensated fields Proc. World Congress
of Medical Physics and the AAPM Annual Congress, August 2000 paper MO-CXH14
Welch M E and Harlow S 2001 The effect of having one multileaf collimator in a
department where it is used for conformal fields Radiother. Oncol. 61 (Suppl. 1) S80
Welsh J, Mehta M, Olivera G, Forrest L, Forouzannia A, Kapatoes J, Ruchala K, Lu W,
Reckwerdt P and Mackie T 2002b Helical tomotherapy: an innovative technology for
delivery of intensity modulated radiotherapy Radiother. Oncol. 64 (Suppl. 1) S211
Welsh J, Olivera G, Hui S, Henderson D, Forouzannia A, Kapatoes J, Ruchala K, Lu W,
Reckwerdt P and Mackie R 2002a Helical tomotherapy with conformal avoidance
appears superior to conventional radiotherapy, 3-D CRT, and IMRT for treatment of
complex tumor volumes Radiother. Oncol. 64 (Suppl. 1) S124
Wierzbicki J and Blackmore L 2003 Comparison of IMRT plans for different delivery
systems Proc. World Congress in Medical Physics (Sydney) CDROM and website
only
Wiesmeyer M D and Beavis A W 2003 A method for using biological imaging data to
modulate equivalent uniform dose based inverse planning Med. Phys. 30 1335
Wilks R J and Bliss P 2002 The use of a compensator library to reduce dose inhomogeneity
in tangential radiotherapy of the breast Radiother. Oncol. 62 14757
Williams P 2002a The practical implementation of IMRT Proc. IPEM Biennial Meeting
(Southampton, 1617 July) 2002 pp 89
2002b Image guided radiotherapyThe next step in conformal radiotherapy
458
References
2003a Integration of imaging and radiation therapy on the linear accelerator Phys.
Medica. 19 47
2003b IMRT: delivery techniques and quality assurance Brit. J. Radiol 76 76676
Williams P C and Cooper P 2000 High-resolution field shaping using a masked multileaf
collimator Phys. Med. Biol. 45 231329
Williams P C and Hounsell A R 2001 X-ray leakage considerations for IMRT Brit. J.
Radiol. 74 98100
Willoughby T R, Weinhous M S and Kupelian P A 2000 Daily transabdominal ultrasound
localization of the prostate gland with the BAT system: 2705 alignments from 100
patients Proc. 42nd Annual ASTRO Meeting. Int. J. Radiat. Oncol. Biol. Phys. 48
(Suppl.) 308
Wilson E, Williams J, Lyn E and Aird E 2001 Active breathing control (ABC) in
the treatment of non-small cell lung cancer (NSCLC) with chart week end less
(CHARTWEL) Radiother. Oncol. 61 (Suppl. 1) S63
Wilson E M, Williams F J, Lyn B E, Wong J J and Aird E G A 2004 Validation of
active breathing control in patients with non-small-cell lung cancer to be treated with
CHARTWEL Int. J. Radiat. Oncol. Biol. Phys. 57 86474
Wong E, Chen J Z and Greenland J 2002a Intensity-modulated arc therapy simplified Int.
Wong J 2000 Prospects of PET imaging in radiation oncology Radiother. Oncol. 56 (Suppl.
1) S7
2001 Quality assurance and verification of IMRT. Phys. Medica. 17 160
2002 Methods to reduce the effects of respirator motion in radiation treatment
2003 Methods to manage respiratory motion in radiation treatment IMRT- the state of
the art: AAPM Medical Physics Monograph Number 29 ed J R Palta and T R Mackie
Wong J W, Armour E, Oldham M, Jaffray D A, Gu R and Jiang L 2002b An image guided
small animal radiation research platform Radiother. Oncol. 64 (Suppl. 1) S61
Wong J, Jaffray D, Siewerdsen J and Sharpe M 2001 Respiratory gated radiation therapy
with active breathing control Med. Phys. 28 1220
Woo M 2000 Verification of MLC field shapes in IMRT treatment using portal imaging
2000 paper WE-FXH-40
Woo M K, Lightstone A W, Shan G, Kumaraswamy L and Li Y 2003 Automatic
verification of step-and-shoot IMRT field segments using portal imaging Med. Phys.
30 34851
Woo S Y, Augspurger M, Teh B S, Lee A G, Uhl B, Grant W and Butler E B 2002
Conformal intensity modulated radiation therapy for the treatment of primary optic
nerve sheath meningloma Radiother. Oncol. 64 (Suppl. 1) S301
Woo S Y, Grant W III, McGary J E, Teh B S and Butler E B 2003 The evolution of quality
assurance for intensity-modulated radiation therapy (IMRT): sequential tomotherapy
Woudstra E, Heijmen B J M and Storchi P R M 2002 Automated sequential selection of
beam orientations Radiother. Oncol. 64 (Suppl. 1) S10
Woudstra E and Storchi P R M 2000 Constrained treatment planning using sequential beam
selection Phys. Med. Biol. 45 213349
References
459
Wu C, Jeraj R, Lu W and Mackie T R 2004 Fast treatment plan modification with an

over-relaxed Cimmino algorithm Med. Phys. 31 191200
Wu C, Jeraj R and Mackie T R 2003a The method of intercepts in parameter space for
the analysis of local minima caused by dose-volume constraints Phys. Med. Biol. 48
N14957
Wu C, Jeraj R, Olivera G and Mackie T R 2001a Comparison between a deterministic and
a stochastic optimizer in radiotherapy optimization Med. Phys. 28 1204
2002a Re-optimization in adaptive radiotherapy Phys. Med. Biol. 47 318195
Wu Q, Arnfield, Tong S, Wu Y and Mohan R 2000c Dynamic splitting of large intensitymodulated fields Phys. Med. Biol. 45 173140
Wu Q, Djajaputra D, Wu Y, Zhou J, Liu H H and Mohan R 2003b Intensity-modulated
radiotherapy optimisation with gEUD-guided dose-volume objectives Phys. Med.
Biol. 48 27991
Wu Q and Mohan R 2000 Algorithms and functionality of an intensity modulated
radiotherapy optimization system Med. Phys. 27 70111
2001 Multiple local minima traps in IMRT optimization Med. Phys. 28 1260
2002 Multiple local minima in IMRT optimisation based on dose-volume criteria
Med. Phys. 29 151427
Wu Q, Mohan R, Morris M, Tong S and Schmidt-Ulrich R 2001e Simultaneous integrated
boost IMRT of advanced head and neck squamous cell carcinomas using dynamic
multi-leaf collimators Int. J. Radiat. Oncol. Biol. Phys. 51 (Suppl. 1) 180
Wu Q, Mohan R and Niemierko A 2000a IMRT optimization based on the generalized
equivalent uniform dose (EUD) Proc. World Congress of Medical Physics and the
AAPM Annual Congress, August 2000 paper MO-G313-05
Wu Q, Mohan R, Niemierko A and Schmidt-Ullrich R 2002b Optimization of intensitymodulated radiotherapy plans based on the equivalent uniform dose Int. J. Radiat.
Wu X, Chen D, Hu X, Luan S, Zhang L and Yu C 2001c A new leaf-sequencing algorithm
for intensity-modulated arc therapy Med. Phys. 28 1252
Wu X, Dibiase S J, Gullapalli R and Yu C X 2004 Deformable image registration for use of
magnetic resonance spectroscopy in prostate treatment planning Int. J. Radiat. Oncol.
Biol. Phys. 58 157783
Wu X and Zhu Y 2001a An optimization method for importance factors and beam weights
based on genetic algorithms for radiotherapy treatment planning Phys. Med. Biol. 46
108599
2001b A maximum-entropy method for the planning of conformal radiotherapy Med.
Phys. 28 22416
Wu X, Zhu Y, Dai J and Wang Z 2000b Selection and determination of beamweights
based on genetic algorithms for conformal radiotherapy treatment planning Phys.
Med. Biol. 45 254758
Wu Y, Wu Q and Mohan R 2002c A method of adaptive filtering and retro-mapping of
intensity-modulated beams to improve delivery efficiency Med. Phys. 29 1304
Wu Y, Yan D, Sharpe M, Miller B and Wong J 2001f Implementing multiple static field
delivery for intensity modulated beams Med. Phys. 28 218897
Xia P, Chuang C F and Verhey L J 2002a Communication and sampling rate limitations in
IMRT delivery with a dynamic multileaf collimator system Med. Phys. 29 41223
460
References
Xia P, Fu K K, Wong G W, Akazawa C and Verhey L J 2000a Comparison of treatment

plans involving intensity-modulated radiotherapy for nasopharyngeal carcinoma Int.
Xia P, Hwang A B and Verhey L J 2002c A leaf sequencing algorithm to enlarge treatment
field length in IMRT Med. Phys. 29 9918
Xia P, Pickett B, Vigneault E, Verhey L J and Roach M III 2001 Forward or inversely
planned segmental multileaf collimator IMRT and sequential tomotherapy to treat
multiple dominant intraprostatic lesions of prostate cancer to 90 Gy Int. J. Radiat.
Xia P and Verhey L 1998 Multileaf collimator leaf-sequencing algorithm for intensity
modulated beams with multiple static segments Med. Phys. 25 142434
Xia P, Wong G, Somers J and Verhey L J 2000b Dosimetric considerations in step and shoot
IMRT delivery Proc. World Congress of Medical Physics and the AAPM Annual
Congress, August 2000 paper MO-GBR-04
Xia P, Yu N, Xing L, Sun X and Verhey L J 2004 Investigation of a new objective function
for inverse planning optimisation Proc. 14th Int.Conf. on the Use of Computers in
Xia P, Yu N, Xing L and Verhey L 2002b Investigation of a variable power objective
function for inverse planning optimization in IMRT Med. Phys. 29 1257
Xiao Y, Bednarz G and Galvin J 2000b A sequential optimization technique for forward
treatment planning for IMRT Proc. World Congress of Medical Physics and the
AAPM Annual Congress, August 2000 paper WE-CXH-67
Xiao Y, Censor Y, Michalski D and Galvin J 2002b Intrinsic smooth intensity pattern
for beamlet-based IMRT inverse planning from a simultaneous projection algorithm
Med. Phys. 29 1254
2001 An IMRT optimization approach that minimizes the total MLC segments Med.
Phys. 28 1262
2002a The least-intensity feasible solution for aperture-based inverse planning in
radiation therapy Private Communication
Xiao Y, Galvin J, Hossain M and Valicenti R 2000a An optimized forward-planning
technique for intensity modulated radiation therapy Med. Phys. 27 20939
Xiao Y, Michalski D, Censor Y and Galvin J 2003 Smoothness evaluation of intensity
pattern resulting from simultaneous subgradient projection method for dose-volume
objective satisfaction Med. Phys. 30 1488
Xing L 2001 Computer verification of MLC leaf sequences for IMRT Med. Phys. 28 1268
9
Xing L, Cotrutz C, Hunjan S, Boyer A L, Adalsteinsson E and Spielman D 2002a Inverse
planning for functional image-guided intensity-modulated radiation therapy Phys.
Med. Biol. 47 356778
Xing L, Crooks S, Li J G, Ozhasoglu C, Chen Y, Mehta V, Goffinet D, Beavis A and Boyer
A 2000b Incorporating respiratory motion into the design of intensity maps in IMRT
treatment of breast cancer Proc. 42nd Annual ASTRO Meeting. Int. J. Radiat. Oncol.
Xing L, Hamilton R J, Spelbring D, Pelizzari C A, Chen G T Y and Boyer A L 1998 Fast
iterative algorithms for three-dimensional treatment planning Med. Phys. 25 18459
Xing L, Hunjan S, Cotrutz C, Yang Y, Boyer A, Adalsteinsson E and Spielman D 2002b
Inverse planning for functional imaging-guided IMRT Med. Phys. 29 1285
References
461
Xing L, Hunjan S, Lian J, Yang Y, Shou Z, Cotrutz C, Schreibmann E, Adalsteinsson

E, Spielman D and Boyer A 2004 Toward biologically conformal radiotherapy:
Functional and molecular image-guided intensity modulated radiation therapy Proc.
14th Int. Conf. on the Use of Computers in Radiation Therapy (Seoul, May) pp 369
Xing L and Li J G 2000 Computer verification of fluence map for intensity modulated
Xing L, Li J, Pugachev A, Le Q and Boyer A 2000a Computer-assisted selection of
importance factors in inverse planning Proc. World Congress of Medical Physics and
the AAPM Annual Congress, August 2000 paper FR-B309-01
Xing L, Yang Y, Li J, Song Y, Pawlicki T, Boyer A and Le Q 2003 Dose matching of an
IMRT plan with an electron or 3D conformal treatment plan Proc. World Congress
Xu B, Yan D, Kota K, Brabbins D S and Martinez A A 2000 The influence of inter and
intra-patient rectum variation on treatment planning for prostate cancer radiotherapy
Xu T 2001 An in vivo verification method for fixed gantry IMRT Med. Phys. 28 12934
2002 A new physical compensation technique for intensity modulated radiation
therapy Med. Phys. 29 1305
Xu T, Al-Ghazi M, Chung H and Molloi S 2003 Re-shapeable physical modulator for
intensity modulated radiation therapy: A treatment planning evaluation Med. Phys.
30 1406
Xu T, Shikhaliev P M, Al-Ghazi M and Molloi S 2002 Reshapable physical modulator for
intensity modulated radiation therapy Med. Phys. 29 22228
Yan H 2003 3D correlation analysis between body marker and target motion for breath
gating control Med. Phys. 30 1338
Yan H, Yin F-F, Guan H-Q and Kim J H 2003 AI-guided parameter optimisation in inverse
treatment planning Phys. Med. Biol. 48 356580
Yan H, Yin F-F and Kim J H 2004 Correlation analysis between skin-marker motions and
internal target motions Proc. 14th Int. Conf. on the Use of Computers in Radiation
Yang C, Chen H, Gollamudi S and Raben A 2002 Physics characteristics and dose
comparison for micro multi-leaf collimator and standard multi-leaf collimator in
whole body conformal radiosurgery Med. Phys. 29 1212
Yang Y and Xing L 2003 Incorporating leaf transmission and head scatter corrections into
step and shoot leaf sequences for IMRT Int. J. Radiat. Oncol. Biol. Phys. 55 112134
2002 An algorithm to incorporate leaf transmission and head scatter corrections into
step-and-shoot leaf sequences for IMRT Med. Phys. 29 1336
Yarnold J R, Donovan E, Bleackley N, Reise S, Regan J, Denholm E, Patel S, Ross G, Tait
D and Evans P 2002 Randomised trial of standard 2D radiotherapy (RT) versus 3D
intensity modulated radiotherapy (IMRT) in patients prescribed breast radiotherapy
Yenice K, Hunt M and Amols H 2001 Advantages of intensity modulated stereotatic
radiosurgery using a mini-multileaf compared with static conformal linac
radiosurgery Med. Phys. 28 12567
Yi B Y 2004 To follow AMC: Aperture manoeuvre with controlled breath Proc. 14th Int.
Yi B Y, Chol E K, Lee S W, Ahn S D, Kim J H, Ji Y H and Vahc Y W 2002 Rtp research
tool box (RTB) for IMRT optimization Radiother. Oncol. 64 (Suppl. 1) S210
462
References
Yin F, Guan H, Ratkewicz A and Kim J 2000 A new technique for 3-D dose reconstruction
using IMRT beams Proc. World Congress of Medical Physics and the AAPM Annual
Congress, August 2000 paper TU-GBR-06
Yin F-F, Yan H, Guan H-Q and Kim J-H 2004 Intelligent parameter optimisation technique
in inverse treatment planning Proc. 14th Int. Conf. on the Use of Computers in
Yoda K and Aoki Y 2003 A multiportal compensator system for IMRT delivery Med. Phys.
30 8806
Yorke E 2001 Advantages of IMRT for dose escalation in radiation therapy of lung cancer
Med. Phys. 28 12912
Young C D, Speight J L, Akazawa P F, Pickett B, Verhey L J and Roach M 2000 Improved
conformal coverage of the prostate with an IMRT potency-sparing technique Proc.
Yu C X 1995 Intensity modulated arc therapy with dynamic multileaf collimation: an
alternative to tomotherapy Phys. Med. Biol. 40 143549
Yu C, Earl M and Shepard D 2004 A scheme for optimising hybrid IMRT Proc. 14th Int.
Yu C and Li J 2001 Delivery optimization with combined rotational and fixed-gantry IMRT
Med. Phys. 28 1306
Yu C X, Li X A, Ma L, Chen D, Naqvi S, Shepard D, Sarfaraz M, Holmes T W,
Suntharalingam M and Mansfield C M 2002 Clinical implementation of intensitymodulated arc therapy. Int. J. Radiat. Oncol. Biol. Phys. 53 45363
Yu C X, Li X A, Ma L, Shepard D, Sarfaraz M, Holmes T, Suntharalingam M and
Mansfield C M 2000 Clinical implementation of intensity-modulated arc therapy
Proc. 42nd Annual ASTRO Meeting. Int. J. Radiat. Oncol. Biol. Phys. 48 (Suppl.)
219
Zackrisson B, Karlsson M and Lofroth P 2000 Biological effect of prolonged fraction
delivery time in intensity modulated radiotherapy (IMRT)in vitro simulation and
model evaluation Radiother. Oncol. 56 (Suppl. 1) S195
Zelefsky M J 2003 Tumour control and tolerance of high dose IMRT in the treatment of
prostate cancer Radiother. Oncol. 68 (Suppl. 1) S62
Zelefsky M J, Fuks Z, Happersett L, Lee H J, Ling C C, Burman C M, Hunt M, Wolfe T,
Venkatraman E S, Jackson A, Skwarchuk M and Leibel S A 2000 Clinical experience
with intensity modulated radiation therapy (IMRT) in prostate cancer Radiother.
Oncol. 55 2419
Zelefsky M J, Fuks Z, Hunt M, Yamada Y, Marion C, Ling C C, Amols H, Venkatraman E
S and Leibel S A 2002 High-dose intensity modulated radiation therapy for prostate
cancer: early toxicity and biochemical outcome in 772 patients Int. J Rad. Oncol.
Biol. Phys. 53 111116
Zelefsky M J and Leibel S A 2003 In response to Drs Schulz and Kagan Int. J. Radiat.
Zhang G, Guerrero T, Huang T-C, Fitzpatrick M J, Ibbott G, Lin K-P and Starkschall
G 2004a 3D optical flow method implementation for mapping of 3D anatomical
structure contours across 4D CT data Proc. 14th Int. Conf. on the Use of Computers
in Radiation Therapy (Seoul, May) pp 35660
Zhang T, Jeraj R, Keller H, Lu W, Olivera G H, McNutt T R, Mackie T R and Paliwal B
2004c Treatment plan optimization incorporating respiratory motion Med. Phys. 31
157686
References
463
Zhang T, Keller H, Jeraj R, Mannon R, Welsh J, Patel R, Mehta M, Mackie R and Paliwal
B 2003a Radiation treatment of the targets with respiratory motion using Breathing
Synchronized Delivery (BSD)A feasibility study Radiother. Oncol. 68 (Suppl. 1)
S745
Zhang T, Keller H, OBrien M J, Mackie T R and Paliwal B 2003b Application of the
spirometer in respiratory gated radiotherapy Med. Phys. 30 316571
Zhang X, Liu H, Wang X, Dong L, Wu Q and Mohan R 2004b Speed and convergence
properties of gradient algorithms for optimisation of IMRT Med. Phys. 31 114152
Zheng Z, Cardarelli G, Shearer D, Wazer D, DiPetrillo T and Chougule P 2002a Peregrine
as a verification tool of Corvus treatment planning system Med. Phys. 29 12312
Zheng Z, Cardarelli G, Shearer D, Wazer D, DiPetrillo T, Chougule P, Puthawala Y
2002b Dosimetry reproducibility of intensity-modulated radiation therapy using a
commercially available system Med. Phys. 29 1263
Zietman A L 2002 Dose escalation in localized prostate cancer: make no assumptions Int.
Zijp L, Sonke J-J and van Herk M 2004 Extraction of the respiratory signal from sequential
thorax cone-beam x-ray images Proc. 14th Int. Conf. on the Use of Computers in
Zinkin H D, Rivard M J, Mignano J E and Wazer D E 2004 Analysis of dose conformity
and normal-tissue sparing using two different IMRT prescription methodologies for
irregularly shaped CNS lesions irradiated with the Beak and 1-cm MIMiC collimators
Zurlo A 2002 Regarding A treatment planning comparison of 3D conformal therapy,
intensity modulated photon therapy and proton therapy for treatment of advanced
head and neck tumours Radiother. Oncol. 63 231
Zygmanski P and Kung J H 2001 Method of identifying dynamic multileaf collimator
irradiation that is highly sensitive to a systematic MLC calibration error Med. Phys.
28 22206
Zygmanski P, Kung J H and Jiang S B 2001b Dependence of IMRT fluence pattern on
DMLC positional tolerance and root-mean-square values (RMS) Med. Phys. 28 1208
Zygmanski P, Kung J, Jiang S and Chen G 2001a Measured dose errors in IMRT delivered
with DMLC in the presence of organ motion: A phantom study Med. Phys. 28 1283
4
2001c A novel method for calculating 3D dose error in IMRT treatment of lung
cancers with and without respiratory gating Med. Phys. 28 1203
Zygmanski P, Kung J, Jiang S and Chin L 2003b Dependence of fluence errors in dynamic
IMRT on leaf-positional errors varying with time and leaf number Med. Phys. 30
273649
Zygmanski P, Kung J H, Jiang S B, Choi N and Chen G 2001d Is IMRT of lung possible
without respiratory gating? Clinical evaluation of a true dose received by a moving
target volume Int. J. Radiat. Oncol. Biol. Phys. 51 (Suppl. 1) 27
Zygmanski P, Kung J, Kooy H, Urribarri J, Bues M and Chen G 2000 A figure of merit
indicating potential dose error in IMRT from inaccurate MLC calibration Proc. World
WE-E313-04
Zygmanski P, Que W, Chen G and Kung J 2003a DMLC sequence that are least sensitive
to intrafractional organ motion Med. Phys. 30 1493

Contemporary IMRT PDF

Uploaded by

Copyright:

Available Formats

Contemporary IMRT PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Contemporary IMRT PDF

Uploaded by

Copyright:

Available Formats

Series in Medical Physics and Biomedical Engineering

Institute of Physics Publishing

Copyright 2005 IOP Publishing Ltd.

c IOP Publishing Ltd 2005

Copyright 2005 IOP Publishing Ltd.

Copyright 2005 IOP Publishing Ltd.

in the Federation. In cooperation with its regional conferences, IFMBE publishes

To organize international cooperation in medical physics in all its aspects,

Copyright 2005 IOP Publishing Ltd.

Preface and acknowledgments

Intensity-modulated radiation therapy (IMRT): General statements

Developments in rotation IMRT and tomotherapy

Developments in IMRT using a multileaf collimator (MLC) (physics)

Copyright 2005 IOP Publishing Ltd.

The best interpreter ever?

Copyright 2005 IOP Publishing Ltd.

3.11.4 Radionics microMLC

Developments in IMRT not using an MLC

Copyright 2005 IOP Publishing Ltd.

Copyright 2005 IOP Publishing Ltd.

Use of contrast agents

Copyright 2005 IOP Publishing Ltd.

Aperture-based planning at the University of Ghent

Copyright 2005 IOP Publishing Ltd.

Copyright 2005 IOP Publishing Ltd.

Preface and acknowledgments

Intensity-modulated radiation therapy (IMRT), virtually non-existent 20 years

Copyright 2005 IOP Publishing Ltd.

Copyright 2005 IOP Publishing Ltd.

1.1 Observations on IMRT at the current time

Copyright 2005 IOP Publishing Ltd.

Intensity-modulated radiation therapy (IMRT)

Figure 1.1. The principle of conformational radiotherapy as exemplified in the historical

for high-spatial-resolution fluence variations, these Japanese developments are

Copyright 2005 IOP Publishing Ltd.

Observations on IMRT at the current time

Traditional radiotherapy with 4

Copyright 2005 IOP Publishing Ltd.

Intensity-modulated radiation therapy (IMRT)

The IMRT Cooperative Working Group generated a lengthy review of IMRT

Copyright 2005 IOP Publishing Ltd.

Observations on IMRT at the current time

The x ray was discovered on 8th November in Germany by Rontgen.

Copyright 2005 IOP Publishing Ltd.

Intensity-modulated radiation therapy (IMRT)

Copyright 2005 IOP Publishing Ltd.

Observations on IMRT at the current time

Copyright 2005 IOP Publishing Ltd.

Intensity-modulated radiation therapy (IMRT)

might open up IMRT for departments or countries without an adequate supply of

Copyright 2005 IOP Publishing Ltd.

Observations on IMRT at the current time

Copyright 2005 IOP Publishing Ltd.

Intensity-modulated radiation therapy (IMRT)

Conversely, Starkshall believes new biological warfare on cancer will make

the early beginnings of IMRT;

1.2 Criticism of the philosophy of IMRT

that there are little data to support general non-investigative use;

Copyright 2005 IOP Publishing Ltd.