Joint CI-JAI advanced accelerator lecture series

Imaging and detectors for medical

physics
Lecture 1: Medical imaging

Dr Barbara Camanzi
barbara.camanzi@stfc.ac.uk
 Course layout

Day AM 09.30 – 11.00 PM 15.30 – 17.00

Week 1
6th June Lecture 1: Introduction to Lecture 2: Detectors for
medical imaging medical imaging
7th June Lecture 3: X-ray imaging
8th June Tutorial
Week 2
13th June Lecture 4: Radionuclides
14th June Lecture 5: Gamma Lecture 6: SPECT
cameras
16th June Lecture 7: PET
Week 3
22nd June Tutorial

Page 2/29
 Books
1. N Barrie Smith & A Webb
Introduction to Medical Imaging
Cambridge University Press

2. Edited by M A Flower
Webb’s Physics of Medical Imaging
CRC Press

3. A Del Guerra
Ionizing Radiation Detectors for Medical Imaging
World Scientific

4. W R Leo
Techniques for Nuclear and Particle Physics Experiments
Springer-Verlag

Page 3/29
 Medical imaging: what is it?
• “Medical imaging is the technique and process of
creating visual representations of the interior of a
body for clinical analysis and medical intervention,
as well as visual representation of the function of
some organs or tissues.” – Wikipedia
• Used to:
1. Diagnose the disease = diagnostic imaging
2. Plan and monitor the treatment of the disease
• Clinical speciality: radiology & radiography +
medical physics

Page 4/29
 Origins of medical imaging
Ref. 2 – Chapter 1

Taken from Ref. 2 pg. 8

Page 5/29
 Origins of medical imaging:
some dates
Ref. 2 – Chapter 1
• 1895: X-ray discovery, Wilhelm C Röntgen
• 1931: invention of cyclotron, Ernest O Lawrence
• 1938: production of 99𝑇𝑐 𝑚 at cyclotron
• 1946: radioisotopes available for public distribution → nuclear
medicine
• 1946: discovery of NMR, Felix Bloch and Edward M Purcell
• 1958 – 1967: SPECT images
• Early 1960s: first clinical PET images
• 1972: first CT machine, Godfrey N Hounsfield
• 1976: first MRI images
• 1978: first commercial SPECT systems
Page 6/29
 Medical imaging techniques
• Techniques not using ionising radiation:
1. Ultrasound imaging (Ref. 1 – Chapter 4, Ref. 2 – Chapter 6)
2. MRI (Ref. 1 – Chapter 5, Ref. 2 – Chapter 7)
3. Infrared imaging (Ref. 2 – Chapter 8)
4. Optical imaging (Ref. 2, Chapter 10)

• Techniques using ionising radiation:

1. X-ray imaging: films, CT
2. SPECT
3. PET
4. Scintigraphy (Ref. 1 – Chapter 3)
Page 7/29
 Basic physics concepts
See for ex. Ref. 4 – Chapters 1, 2, 3 and 4

• X-ray energy spectrum

• Interaction of photons with matter
1. Photoelectric effect
2. Compton scattering
3. Absorption and attenuation
• Probability distributions: Gaussian, Poisson, etc.
• Radioactive sources
1. Radioactive decay
• Dosimetric units
Page 8/29
 The receiver operating
characteristic (ROC) curve
True positive fraction
Diagnosis Actual situation 1
Disease Healthy
Disease True False
positive positive
Healthy False True
negative negative
0.5 Area under the curve
𝑵𝒄𝒐𝒓𝒓𝒆𝒄𝒕 𝒅𝒊𝒂𝒈𝒏𝒐𝒔𝒆𝒔 measures quality of
𝑨𝒄𝒄𝒖𝒓𝒂𝒄𝒚 = diagnostic procedure
𝑵𝒕𝒐𝒕𝒂𝒍 𝒅𝒊𝒂𝒈𝒏𝒐𝒔𝒆𝒔

𝑵𝒕𝒓𝒖𝒆 𝒑𝒐𝒔𝒊𝒕𝒊𝒗𝒆𝒔
𝑺𝒆𝒏𝒔𝒊𝒕𝒊𝒗𝒊𝒕𝒚 =
𝑵𝒕𝒓𝒖𝒆 𝒑𝒐𝒔𝒊𝒕𝒊𝒗𝒆𝒔 + 𝑵𝒇𝒂𝒍𝒔𝒆 𝒏𝒆𝒈𝒂𝒕𝒊𝒗𝒆𝒔
0
𝑵𝒕𝒓𝒖𝒆 𝒏𝒆𝒈𝒂𝒕𝒊𝒗𝒆𝒔
𝑺𝒑𝒆𝒄𝒊𝒇𝒊𝒄𝒊𝒕𝒚 = 0 0.5 1
𝑵𝒕𝒓𝒖𝒆 𝒏𝒆𝒈𝒂𝒗𝒊𝒕𝒆𝒔 + 𝑵𝒇𝒂𝒍𝒔𝒆 𝒑𝒐𝒔𝒊𝒕𝒊𝒗𝒆𝒔 False positive fraction
Page 9/29
 Exercise
True positive fraction

Exercise 1
Draw the ROC curve for when
trying to diagnose cardiac
disease by counting the number
of lesions in the brain
Solution 0.5

− Cardiac disease completely

unrelated to brain lesions
− 50-50 chance of true positives
and false positives irrespective
of number of brain lesions 0
found 0 0.5 1
False positive fraction
Page 10/29
 Important parameters for all
imaging techniques
• Spatial resolution
• Signal-to-noise ratio, SNR
• Contrast-to-noise ratio, CNR

Page 11/29
 Spatial resolution
• Imaging systems not perfect → introduce blurring
= no sharp edges → finite spatial resolution
• Spatial resolution determines:
1. The smallest feature that can be visualised
2. The smallest distance between two features so that
they can be resolved and not seen as one
• Measures of spatial resolution / blurring:
1. Line spread function (LSF) – 1D
2. Point spread function (PSF) – 3D

Page 12/29
 Line spread function (LSF)
• Measured by imaging a single thin line or set of lines

Taken from Ref. 1 pg. 6

• For many imaging systems 𝐿𝑆𝐹 is a Gaussian:

1 𝑦 − 𝑦0 2
𝐿𝑆𝐹 𝑦 = 𝑒𝑥𝑝 − 2
2𝜋𝜎 2 2𝜎
𝜎 = standard deviation
𝐹𝑊𝐻𝑀 = 2 2 ln 2 𝜎 ≅ 2.36𝜎
Page 13/29
𝐿𝑆𝐹 and spatial resolution

Objects
• Two structures can be
resolved if:
𝐿𝑆𝐹1 𝑑 > 𝐹𝑊𝐻𝑀𝐿𝑆𝐹 𝑑 > 2.36𝜎𝐿𝑆𝐹
𝐹𝑊𝐻𝑀𝐿𝑆𝐹1
𝑑 = distance between two
structures
𝐿𝑆𝐹2 • The narrower 𝐿𝑆𝐹 = smaller
𝐹𝑊𝐻𝑀𝐿𝑆𝐹2 > 𝐹𝑊𝐻𝑀𝐿𝑆𝐹1 𝐹𝑊𝐻𝑀 → the smaller the
distance between two
structures that can be
𝐿𝑆𝐹3 resolved → better spatial
𝐹𝑊𝐻𝑀𝐿𝑆𝐹3 > 𝐹𝑊𝐻𝑀𝐿𝑆𝐹2
resolution

Page 14/29
 Point spread function (PSF)
• Takes into account the spatial resolution may
become poorer with depth in the body → 3D
• PSF describes image of a point source
Taken from Ref. 1 pg. 8

Worsening PDF

𝐼 𝑥, 𝑦, 𝑧 = 𝑂 𝑥, 𝑦, 𝑧 ∗ ℎ 𝑥, 𝑦, 𝑧
3D image 3D object convolution 3D PSF
Page 15/29
 Signal-to-noise ratio SNR
• Noise = random signal superimposed on top of
real signal → mean value zero → standard
deviation 𝜎𝑁
• Sources of noise different for different imaging
modalities
• Signal-to-noise ratio 𝑆𝑁𝑅:
𝑆𝑖𝑔𝑛𝑎𝑙
𝑆𝑁𝑅 =
𝜎𝑁
• The higher 𝑆𝑁𝑅 the better the image:
– Maximise signal = when designing imaging systems
– Average signal acquired over repeated scans
Page 16/29
Example of effect of noise:
MRI scan

Taken from Ref. 1 pg. 11

Page 17/29
 Example of averaging the signal:
MRI scan

Taken from Ref. 1 pg. 11

a. One scan
b. Average of two identical scans
c. Average of four identical scans
d. Average of 16 identical scans
Page 18/29
 Contrast-to-noise ration CNR
• Ability to distinguish between different tissues =
between healthy and pathological tissues
• Contrast-to-noise ratio 𝐶𝑁𝑅𝐴𝐵 between tissues 𝐴
and 𝐵:
𝐶𝐴𝐵 𝑆𝐴 − 𝑆𝐵
𝐶𝑁𝑅𝐴𝐵 = = = 𝑆𝑁𝑅𝐴 − 𝑆𝑁𝑅𝐵
𝜎𝑁 𝜎𝑁
𝑆𝐴 , 𝑆𝐵 = signals from tissues 𝐴 and 𝐵
𝐶𝐴𝐵 = contrast between tissues 𝐴 and 𝐵
𝜎𝑁 = standard deviation of noise
• The higher 𝐶𝑁𝑅𝐴𝐵 the better the image
Page 19/29
 Dose
𝑟𝑎𝑑𝑖𝑎𝑡𝑖𝑜𝑛 𝑒𝑛𝑒𝑟𝑔𝑦 𝐸(𝐽)
• 𝐴𝑏𝑠𝑜𝑟𝑏𝑒𝑑 𝑑𝑜𝑠𝑒 𝐷 𝐺𝑦 =
𝑘𝑔 𝑜𝑓 𝑡𝑖𝑠𝑠𝑢𝑒
• 𝑀𝑒𝑎𝑛 𝑎𝑏𝑠𝑜𝑟𝑏𝑒𝑑 𝑑𝑜𝑠𝑒 𝐷𝑇,𝑅 𝑖𝑛 𝑚𝑎𝑠𝑠 𝑚 𝑇 𝑜𝑓 𝑡𝑖𝑠𝑠𝑢𝑒 𝑇
𝑓𝑟𝑜𝑚 𝑎𝑚𝑜𝑢𝑛𝑡 𝑜𝑓 𝑟𝑎𝑑𝑖𝑎𝑡𝑖𝑜𝑛 𝑅
1
𝐷𝑇,𝑅 = 𝐷𝑅 𝑑𝑚
𝑚𝑇
𝑚𝑇
• 𝐸𝑞𝑢𝑖𝑣𝑎𝑙𝑒𝑛𝑡 𝑑𝑜𝑠𝑒 𝐻𝑇 (𝑆𝑣) = 𝑅 𝑤𝑅 𝐷𝑇,𝑅
with 𝑤𝑅 𝑟𝑎𝑑𝑖𝑎𝑡𝑖𝑜𝑛 𝑤𝑒𝑖𝑔ℎ𝑡𝑖𝑛𝑔 𝑓𝑎𝑐𝑡𝑜𝑟

Page 20/29
 Dose damage

Radiation effects Description Probability

Deterministic Cellular damage that leads to - Dose threshold1
loss in tissue function - Zero or low at low
doses below threshold
- Climbs rapidly to unity
above threshold
Stochastic Cells are not killed but undergo - No dose threshold
genetic mutations → cancer - Increases with the
dose
- In the low dose region
non negligible and
higher than for
deterministic effects
1The doses in imaging procedures are usually below this threshold

Page 21/29
 Effective dose
• Some tissues are more sensitive to radiation dose
than others
• Tissue weighting factor 𝑤𝑇 = fraction of total
stochastic radiation risk
𝑇 𝑤𝑇 = 1

• 𝐸𝑓𝑓𝑒𝑐𝑡𝑖𝑣𝑒 𝑑𝑜𝑠𝑒 𝐸 = 𝑇 𝑤𝑇 𝐻𝑇

Page 22/29
 Tissue weighting factors
Tissue / organ Tissue weighting factor1
Gonads 0.2
Bone marrow (red) 0.12
Colon 0.12
Lung 0.12
Stomach 0.12
Chest 0.05
Bladder 0.05
Liver 0.05
Thyroid 0.05
Oesophagus 0.05
Average (brain, small intestines, adrenals, kidney, pancreas, muscle, spleen, 0.05
thymus, uterus)
Skin 0.01
Bone surface 0.01
1 For the gonads = risk of hereditary conditions, for all others organs = risk of cancer

Page 23/29
 Multimodality imaging
Ref. 2 – Chapter 15

• Multimodality imaging (MMI) obtains information

from combination of image data
• Main use = visualisation = localise functional data
within body anatomy for same patient
• Made possible by:
1. Increasing computer power easily accessible
2. Merging different imaging modalities: hybrid scanners
3. Multidisciplinary teams

Page 24/29
 MMI: image registration
• Image registration = align image data sets to achieve
spatial correspondence for direct comparison
– Images from different modalities
– Images from same modality at different times
– Images with standardised anatomy’s atlases

1. Images produced already aligned → nothing to be done

– Set-up parameters identical for all scans
2. Images not already aligned → image transformation
– Rigid-body registration
– Non-rigid registration
Page 25/29
 MMI: rigid-body registration
• Image transformation:
𝑝𝑖′ = 𝑹 𝑺𝒑′′
𝒊 + 𝑻 + 𝑖 𝑖 = 1 − 𝑁

𝑝𝑖′ , 𝑝𝑖′′ = two 3D point sets

𝑹 = 3x3 rotation matrix
𝑺 = 3x3 scaling matrix (diagonal)
𝑻 = 3x1 translation vector
 = 3x1 ‘noise’ or ‘uncertainty’ vector
• 9 parameters to be determined:
3 scaling factors + 3 angles + 3 translation factors
Page 26/29
 MMI: Classification schemes
Mainly used with rigid-body registration:
1. Point Matching
− Based on a landmark that can be considered as point
− Minimum 3 non-coplanar points needed, accuracy increases with
number of points
− Anatomical points accurate, functional data less accurate
2. Line or Surface Matching
− Lines or Surfaces obtained from external frames or internal anatomy
3. Volume Matching
− Derived from voxel information
− Various algorithms being developed

Page 27/29
 MMI: non-rigid-body registration
See for ex. W R Crum, T Hartkens and D L G Hill “Non-rigid image registration:
theory and practice”, Brit. Journ. Rad. 77 (2004), S140–S153

• Anatomy not rigid (apart some bony structures)

→ errors
• Elastic registration takes into account:
1. Movement between different scans
2. Motion during the scan
• Various methods developed
• Less commonly used due to high complexity and
pitfalls of warping image data unrealistically

Page 28/29
 MMI registration accuracy
• Aim = one-to-one spatial correspondence
between each image’s element
• Potential primary sources of errors:
1. Definition of features used
2. Accuracy and robustness of algorithm used
3. Accuracy of image transformation
4. Differences in anatomy or image parameters
• Absolute accuracy impossible to determine, only
relative accuracy between different registration
methods

Page 29/29