Get MRI of The Liver A Practical Guide 1st Edition J.A. Robinson Philip Free All Chapters
Get MRI of The Liver A Practical Guide 1st Edition J.A. Robinson Philip Free All Chapters
Get MRI of The Liver A Practical Guide 1st Edition J.A. Robinson Philip Free All Chapters
com
https://ebookname.com/product/mri-of-the-liver-a-practical-
guide-1st-edition-j-a-robinson-philip/
OR CLICK BUTTON
DOWLOAD NOW
https://ebookname.com/product/practical-field-ecology-a-project-
guide-1st-edition-c-philip-wheater/
https://ebookname.com/product/model-based-predictive-control-a-
practical-approach-control-series-1st-edition-j-a-rossiter/
https://ebookname.com/product/drugs-and-the-liver-a-guide-to-
drug-handling-in-liver-dysfunction-1st-edition-penny-north-lewis/
https://ebookname.com/product/practical-management-of-liver-
diseases-1st-edition-zobair-m-younossi/
ADHD in Adults A Practical Guide to Evaluation and
Management 1st Edition Philip Asherson M.B.
https://ebookname.com/product/adhd-in-adults-a-practical-guide-
to-evaluation-and-management-1st-edition-philip-asherson-m-b/
https://ebookname.com/product/practical-audio-electronics-1st-
edition-kevin-robinson/
https://ebookname.com/product/a-clinician-s-guide-to-rheumatic-
diseases-in-children-1st-edition-thomas-j-a-lehman/
https://ebookname.com/product/doing-right-a-practical-guide-to-
ethics-for-medical-trainees-and-physicians-philip-c-hebert/
https://ebookname.com/product/a-practical-guide-to-the-histology-
of-the-mouse-1st-edition-scudamore/
MRI
of the Liver
A Practical Guide
Philip J. A. Robinson
University of Leeds
St. James's University Hospital
Leeds, U.K.
Janice Ward
St. James's University Hospital
Leeds, U.K.
This book contains information obtained from authentic and highly regarded sources. Reprinted material is quoted with permission, and sources are
indicated. A wide variety of references are listed. Reasonable efforts have been made to publish reliable data and information, but the author and the
publisher cannot assume responsibility for the validity of all materials or for the consequences of their use.
No part of this book may be reprinted, reproduced, transmitted, or utilized in any form by any electronic, mechanical, or other means, now known
or hereafter invented, including photocopying, microfilming, and recording, or in any information storage or retrieval system, without written
permission from the publishers.
For permission to photocopy or use material electronically from this work, please access www.copyright.com (http://www.copyright.com/) or contact
the Copyright Clearance Center, Inc. (CCC) 222 Rosewood Drive, Danvers, MA 01923, 978-750-8400. CCC is a not-for-profit organization that
provides licenses and registration for a variety of users. For organizations that have been granted a photocopy license by the CCC, a separate system
of payment has been arranged.
Trademark Notice: Product or corporate names may be trademarks or registered trademarks, and are used only for identification and explanation
without intent to infringe.
MRI of the liver has come of age. After a prolonged gradual period of development,
MRI is now sufficiently accessible to radiologists with an interest in hepatobiliary
imaging, and MRI techniques are sufficiently stable for this practical guide to be of
value. We have been privileged to obtain extensive experience of liver MRI over the last
15 years, so committing to paper the knowledge, ideas, and opinions that we have
accumulated during that time has been a natural consequence of our work.
This book is aimed at clinical professionals—from any discipline—who wish to
develop or to supplement an interest in the clinical applications of MRI in patients
with liver disease. After the first four chapters, which relate to the clinical indications,
techniques, contrast agents, and anatomic aspects of liver MRI, we have taken a broadly
systematic approach to the recognition and interpretation of different liver disorders.
For us it is axiomatic that by the time a patient is referred for MRI of the liver, he or
she will already have undergone first line investigations which typically include sono-
graphy and liver function tests. Each of the Chapters 5–13 are based upon the outcome
of preliminary liver screening results—for example, cysts and cyst-like lesions are dealt
with in Chapter 5, solid focal lesions in the non-cirrhotic liver in Chapter 6 and so on.
One of our major aims in producing this book has been to offer the reader a large
number of illustrative examples of different pathologies. We have deliberately included
multiple examples of some pathologies in order to indicate the potential range of
appearances, including atypical as well as typical cases. Essential technical aspects of
image acquisition are included in each of the “clinical” chapters, but the reader is urged
to refer to Chapters 2 and 3 for a more detailed discussion of the choice of acquisition
sequences and the use of contrast agents. Wherever possible we have also included
a discussion of the differential diagnosis of conditions that may appear similar,
and illustrative examples have been provided. This has inevitably led to some overlap
between the contents of different chapters, but it has also allowed us to reduce the
frequency of cross-referencing between chapters.
Well over 99% of the images illustrated here are from our own department, so
we have to admit that this book presents a personal and possibly idiosyncratic view of
our subject. However, presenting personal experience does allow us to be more detailed
and more specific in our recommendations for technique and interpretation, and we
hope this aspect will improve the practical value of this guide to the less experienced
reader. For the same reasons we have not provided exhaustive references to support all
the points made in the text, but instead offer a limited selected bibliography.
Finally, this book is aimed at clinical professionals who share our view that the
role of the investigating diagnostician is not just to interpret a set of images put in front
of her or him, but to establish what is wrong with the patient. As mentioned above,
iii
iv PREFACE
when a patient presents for liver MRI the results of the first line investigations should
have crystallized the clinical question, and a careful review of the history and results
of initial investigations is often helpful in guiding the interpretion of the MRI findings
and assessing their significance. Particularly in patients who have already undergone
treatment, it is often critical to establish the details—has the patient had previous liver
surgery? if so, exactly what procedure was carried out and what is the expected residual
anatomy? has the patient been treated with chemotherapy? has the patient had thermal
ablation of liver lesions, and if so, how long ago? The more you know when you start,
the easier it is to interpret the result of further imaging.
There is a second reason for carrying out a thorough assessment of each patient
before undertaking liver MRI—that is to ensure the correct technical approach. We need
to establish what exactly we are trying to discover, in order to ensure the appropriate
choice of contrast agents and to select the relevant acquisition sequences. In our view,
liver MRI is not a “one size fits all” technique—each patient must be assessed indivi-
dually. This is not to say that there is an infinite range of technical approaches—most
patients can be successfully investigated by selecting from a short menu of protocols,
each comprising a set of sequences with one or more contrast agents. Obtaining good
results with liver MRI are only achieved with careful attention to practical details, and a
thorough study of the clinical context.
This volume could not have been created without the help and support of our staff
and colleagues at St. James’s University Hospital. We have enjoyed enthusiastic and
productive collaboration with colleagues in hepatology, liver surgery, and pathology.
We have had the continuing support of an enthusiastic group of radiographers and
radiographic assistants in the MR section. We are grateful to our radiologic colleagues
for their co-operation and support, particularly Dr. M.B. Sheridan and Dr. H. Woodley.
Our regular collaborator Dr. J. Ashley Guthrie has generously allowed us to use a large
number of his cases. Dr. Daniel Wilson has given us sustained scientific support and
much helpful discussion. We would like to thank Mrs. S. Callaghan for secretarial
assistance and we are particularly indebted to Mrs. S. Boyes who has spent many hours
retrieving images and preparing them for publication, in addition to her help with
references and typing.
Philip J. A. Robinson
Janice Ward
Contents
Preface . . . . iii
v
vi CONTENTS
Index . . . . 393
PART A
1
CHAPTER 1
3
4 PART A: WHY, WHEN, AND HOW TO PERFORM MRI OF THE LIVER
In order to clarify the role of different imaging techniques in investigation of liver disease,
we need first to answer two questions:
In terms of 3-dimensional spatial resolution, when imaging an organ as large as the liver
in ideal conditions there is little to choose between sonography, CT, and MRI. The major
differences are in intrinsic contrast resolution and in temporal resolution. The advantage
CHAPTER 1: ROLE OF MRI IN LIVER DISEASE 5
2. Surgical Cases—Patients who are candidates for liver resection, for transplantation or for
non-surgical local therapies (ablation or chemoembolization) of tumors.
The last decade has seen a dramatic increase in the use of surgical resection and various
ablation techniques for the treatment of malignant liver tumors. In selecting patients for
treatment, we need to use the most sensitive technique available for detecting small
metastases, and differentiating them from small benign liver lesions. Pre-operative
mapping to show the relationships of tumors to the main vascular structures, and the
anatomy of the portal circulation in cirrhotic patients who are candidates for liver
transplantation, is also conveniently incorporated into the MR examination.
(a) Like other workers in the field, we have found liver MRI to be more sensitive than
multislice CT in detecting small metastases, particularly from colorectal cancer (1–13).
(b) Like other workers in the field, we have found liver MRI to be more specific in
characterizing various benign pathologies presenting as focal abnormalities within the
liver (6,7,10,12–17).
(c) Like other workers in the field, we have found MRI to be the most effective technique for
the early detection and characterization of hepatocellular cancer in cirrhotic patients
(18–24).
However, it must be emphasised that MRI of the liver is not a “plug, play, and walk
away” technique. Obtaining good results requires consideration of the clinical question in
each individual patient, appropriate choice of techniques, and contrast agents, and
meticulous attention to detail in the acquisition and interpretation of the images.
Technical developments in MRI continue apace, and it is inevitable that the snapshot
provided above will change in coming years. The introduction of parallel imaging and the
use of 3T field strength are examples of the new technology which will no doubt impact on
hepatobiliary applications of MRI in the near future. We can look forward to further
improvements.
REFERENCES
1. Hagspiel KD, Neidl KF, Eichenberger AC, Weder W, Marincek B. Detection of liver metastases:
comparison of superparamagnetic iron oxide-enhanced and unenhanced MR imaging at 1.5T with
dynamic CT, intraoperative US, and percutaneous US. Radiology 1995; 196:471–478.
2. Huppertz A, Haraida S, Kraus A, et al. Enhancement of focal liver lesions at gadoxetic acid-enhanced MR
imaging: correlation with histopathologic findings and spiral CT-initial observations. Radiology 2005;
243:468–478.
3. Kondo H, Kanematsu M, Hoshi H, et al. Preoperative detection of malignant tumors: comparison of
combined methods of MR imaging with combined methods of CT. Am J Roentgenol 2000; 174:947–954.
4. Lencioni R, Della Pina C, Bruix J, et al. Clinical management of hepatic malignancies: ferucarbotran-
enhanced magnetic resonance imaging versus contrast-enhanced spiral computed tomography. Dig Dis
Sci 2005; 50:533–537.
5. Muller RD, Vogel K, Neumann K, et al. SPIO-MR imaging versus double-phase spiral CT in detecting
malignant lesions of the liver. Acta Radiol 1999; 40:628–635.
6. Poeckler-Schoeniger C, Koepke J, Gueckel F, et al. MRI with superparamagnetic iron oxide: efficacy in the
detection and characterization of focal hepatic lesions. Magn Reson Imaging 1999; 17:383–392.
CHAPTER 1: ROLE OF MRI IN LIVER DISEASE 7
7. Reimer P, Jähnke N, Fiebich M, et al. Hepatic lesion detection and characterization: value of non-
enahnced MR imaging, superparamagnetic iron oxide-enhanced MR imaging and spiral CT-ROC
analysis. Radiology 2000; 217:152–158.
8. Semelka RC, Cance WG, Marcos HB, Mauro MA. Liver metastases: comparison of current MR techniques
and spiral CT during arterial portography for detection in 20 surgically staged cases. Radiology 1999;
213:86–91.
9. Seneterre E, Taourel P, Bouvier Y, et al. Detection of hepatic metastases: ferumoxides-enhanced MR
imaging versus unenhanced MR imaging and CT during arterial portography. Radiology 1996;
200:785–792.
10. Takahama K, Amano Y, Hayashi H, et al. Detection and characterization of focal liver lesions using
superparamagnetic iron oxide-enhanced magnetic resonance imaging: comparison between ferum-
oxides-enhanced T1-weighted imaging and delayed-phase gadolinium-enhanced T1-weighted imaging.
Abdom Imaging 2003; 28:525–530.
11. Ward J, Guthrie JA, Wilson D, et al. Colorectal hepatic metastases: detection with SPIO-enhanced breath-
hold MR imaging—comparison of optimized sequences. Radiology 2003; 228:709–718.
12. Ward J, Naik KS, Guthrie JA, et al. Hepatic lesion detection: comparison of MR imaging after the
administration of superparamagnetic iron oxide with dual-phase CT by using alternative-free response
receiver operating characteristic analysis. Radiology 1999; 210:459–466.
13. Ward J, Robinson PJ, Guthrie JA, et al. Liver metastases in candidates for hepatic resection: comparison of
helical CT and gadolinium- and SPIO-enhanced MR imaging. Radiology 2005; 237:170–180.
14. Hawighorst H, Schoenberg S, Knopp MV, et al. Hepatic lesions: morphologic and functional
characterization with multiphase breath-hold 3D gadolinium-enhanced MR angiography—initial
results. Radiology 1999; 210:89–96.
15. Ito K, Mitchell DG. Imaging diagnosis of cirrhosis and chronic hepatitis. Intervirology 2004; 47:134–143.
16. Low RN. MR imaging of the liver using gadolinium chelates. In: MR Imaging of the liver 1: techniques
and contrast agents. Magn Reson Imaging Clin N Am 2001:717–743.
17. Ward J, Robinson PJ. Combined use of MR contrast agents for evaluating liver disease. Magn Reson
Imaging Clin N Am 2001; 9:767–802.
18. Burrel M, Llovet JM, Ayuso C, et al. Barcelona clinic liver cancer group. MRI angiography is superior to
helical CT for detection of HCC prior to liver transplantation: an explant correlation. Hepatology 2003;
38:1034–1042.
19. Ichikawa T. MRI in the evaluation of hepatocellular nodules: role of pulse sequences and contrast agents.
Intervirology 2004; 47:252–270.
20. Kwak HS, Lee JM, Kim CS. Preoperative detection of hepatocellular carcinoma: comparison of combined
contrast-enhanced MR imaging and combined CT during arterial portography and CT hepatic
arteriography. Eur Radiol 2004; 14:447–457.
21. Oi H, Murakami T, Kim T, et al. Dynamic MR imaging and early-phase helical CT for detecting small
intrahepatic metastases of hepatocellular carcinoma. Am J Roentgenol 1996; 166:369–374.
22. Rode A, Bancel B, Douek P, et al. Small nodule detection in cirrhotic livers: evaluation with US, spiral CT,
and MR and correlation with pathologic examination of explanted liver. J Comput Assist Tomogr 2001;
25:327–336.
23. Ward J, Guthrie JA, Scott DJ, et al. Hepatocellular carcinoma in the cirrhotic liver: double-contrast MR
imaging for diagnosis. Radiology 2000; 216:154–162.
24. Yamamoto H, Yamashita Y, Yoshimatsu S, et al. Hepatocellular carcinoma in cirrhotic livers: detection
with unenhanced and iron oxide-enhanced MR imaging. Radiology 1995; 195:106–112.
CHAPTER 2
- Susceptibility . . . . 12
- Chemical Shift . . . . 12
First-Order Chemical Shift . . . . 12
Second-Order Chemical Shift . . . . 13
- Fat Suppression . . . . 13
- T2-Weighted Imaging . . . . 13
- Greater T2 Weighting . . . . 14
- STIR Sequences . . . . 15
- Hybrid Sequences . . . . 15
- T2-Weighted GRE . . . . 16
- T1-Weighted Imaging . . . . 16
- Illustrative Figures . . . . 18
- References . . . . 18
9
10 PART A: WHY, WHEN, AND HOW TO PERFORM MRI OF THE LIVER
Imaging speed and spatial resolution are influenced by gradient performance. Both gradient
strength and the time taken to achieve maximum gradient amplitude (rise time) are
important. Recent developments in gradient coil design and technology have resulted in
the routine use of gradients with a maximum strength of 30–40 mT/m, approximately
three times the strength of those previously used (standard gradient strength typically
10–15 mT/m). Gradient rise times have also been substantially reduced to approximately
200 msec (traditionally approximately 600 msec).
Increasing the speed of scanning is generally achieved at the expense of signal to noise
ratio (SNR) but this is largely compensated by the use of body phased array coils.
Refinements in gradient technology have enabled rapid imaging without compromising
SNR. With 2D imaging, high performance gradients can allow a shorter TE (enabling more
slices for a given TR) at a fixed bandwidth, or a narrower bandwidth giving better SNR for
the same TE. With 3D imaging they enable shorter TRs which allow more slices or better
resolution for a given acquisition time. Receiver bandwidth is an important determinant of
SNR. The bandwidth represents the range of frequencies contained within the MR signal and
determines the rate at which an echo is sampled. TE is determined by the duration of the
dephasing lobe of the readout gradient and the length of acquisition, with maximum signal
amplitude of the echo occurring when the amount of rephasing equals dephasing. With a
low bandwidth sequence the echo is sampled at a lower sampling frequency for longer. This
results in high SNR but the center of the echo occurs later, TE is longer and fewer slices are
available for a given TR. TE may be shortened by increasing the gradient amplitude and
sampling at a higher frequency for less time (high bandwidth sequence) but then because
more frequencies are being sampled there is more noise in the resultant image (Fig. 2).
Although the sampling rate is fixed for a particular bandwidth, high performance gradients
enable more rapid sampling at lower bandwidths because the dephasing gradient is applied
at higher amplitude for less time, thus shortening the duration of image acquisition and TE.
Also, because gradient rise time is much faster, the time to maximum signal amplitude
occurs sooner, which further reduces TE while maintaining SNR. Further shortening of the
TE at the same bandwidth may be achieved with the use of asymmetric echoes. By reducing
either the duration or amplitude of the dephasing gradient the echo occurs towards the start
of data acquisition instead of the center (Fig. 3). Although TE is shorter there is an overall
decrease in SNR because only part of the echo is sampled; the degree of asymmetry
determines how much of K space is filled, and the remainder is encoded with zero filling.
Although low bandwidths are recommended for use with GRE sequences, they are generally
not suitable for use with FSE sequences. As bandwidth decreases, echo spacing increases
and this adversely effects image quality. With longer echo spacing, blurring, susceptibility,
and edge-related artifacts are more pronounced, so with most FSE sequences the shortest
possible echo spacing commensurate with adequate SNR is automatically implemented.
Given the typically elliptical shape of the body outline, a rectangular field of view (FOV)
may also be used to shorten acquisition times. With this technique the square raw data
matrix is sampled unequally with fewer data points sampled in the phase encoding
direction. If the increment between each phase-encoding step is doubled, FOV is halved.
Since both the FOV and the number of phase-encoding steps is halved, pixel size in the phase
encoding direction and therefore spatial resolution is unchanged. A 75% FOV is suitable
for most patients but varies depending on patient shape. The technique incurs a small
reduction in SNR because only part of K space is sampled.
CHAPTER 2: TECHNIQUES FOR MRI OF THE LIVER 11
Parallel imaging techniques further reduce scan time or allow higher resolution images to
be obtained in the same time as conventional images (Fig. 4). With parallel imaging,
spatial information from the different receiver channels of a multi-coil array is used to
replace a proportion of the lines of K space and speed up acquisition time. Although the
whole of K space is measured in a conventional fashion, the spacing in between the lines
of K space is increased, so fewer echoes are measured and scan times are reduced without
compromising spatial resolution. With conventional scanning fold-over may occur when
the spacing between the lines of K space is increased to reduce FOV (Fig. 5). Parallel
imaging eliminates wrap artifact either by using the coil sensitivities to “unfold” the
image (e.g., SENSE) or by using the coil sensitivities to generate the missing lines of K
space (e.g., SMASH/GRAPPA) to increase the FOV and eliminate fold-over. All the above
parallel imaging techniques require some estimation of the individual coil sensitivities.
These sensitivity maps are acquired either prior to, or during parallel image acquisition by
acquiring a small number of calibration K space lines during the actual scan. The scan
time in general, is reduced in proportion to the number of coil elements within the multi-
coil array. Good image quality is currently achievable with scan time reductions of !2 to
!4, but further reduction of scan time requires specialized hardware. The decrease in
acquisition time is at the expense of SNR, which is reduced by approximately the square
root of the scan time reduction factor.
allow the correct anatomical alignment of sequential slices when multiple single slices are
obtained during a single acquisition (Fig. 7). They also allow high-resolution non-breathhold
images to be acquired over several minutes without motion artifact (Fig. 8).
Motion-induced ghost artifacts have the same configuration as the moving structure they
arise from, and they always occur in the phase-encoding direction. In most patients
breathholding effectively eliminates artifacts from structures such as the gallbladder or
anterior abdominal wall, but pre-saturation or gradient motion rephasing (GMR) techniques
are necessary to eliminate ghosting from vascular structures (Fig. 9). Flow artifacts may arise
from any vessel but those from the aorta are usually more pronounced. They are readily
identified because they occur at regular intervals along the phase encoding axis. The distance
between the ghosts increases with increasing TR and they are more pronounced at longer TEs
because more time is available for motion to occur. When the phase- and frequency-encoding
directions are exchanged, the orientation of the artifact will also change (Fig. 10). With GMR,
additional gradient lobes added to the standard dephasing and rephasing lobes of the slice
select and readout gradients before the echo is sampled, compensate in advance for motion-
induced phase shifts that occur between excitation and echo sampling. Because more time is
needed to utilise the additional gradient lobes, GMR techniques—although effective—are
only feasible with sequences using relatively longer TE. Use of a higher bandwidth will
reduce the minimum TE feasible with GMR but at the expense of SNR. Regardless of TE, pre-
saturation pulses positioned above and below the region of interest are also effective (Fig. 11).
Their implementation will, however, reduce the number of slices available for a given scan
time with 2D imaging or increase the TR for 3D imaging.
SUSCEPTIBILITY
Susceptibility artifacts are caused by a local distortion of the magnetic field due to the close
proximity of objects or substances (mainly metallic foreign bodies and air) which
are susceptible to the induction of magnetization. The severity of susceptibility artifacts is
pulse sequence dependent. The 1808 refocusing pulse of spin echo sequences reduces
susceptibility by correcting dephasing and this is particularly effective on FSE sequences
with a long echo train, due to the incorporation of multiple 1808 refocusing pulses. Because
of this, FSE sequences are relatively insensitive to this type of artifact. Susceptibility is also
less pronounced at shorter TEs because there is less time for dephasing to occur (Fig. 12).
GRE sequences are particularly sensitive to susceptibility artifact because they have no 1808
refocusing pulse, but the artifact is minimized when very short TE is used.
CHEMICAL SHIFT
There are two types of chemical shift artifact, described as first order (caused by mis-
registration) and second order (caused by phase cancellation).
Fat protons resonate at a slightly lower frequency than water protons (with a difference of
approximately 3.5 parts per million at 1.5T). In tissues where fat and water protons are
adjacent, frequency encoding errors occur because the signal from the fat protons has a
slightly lower frequency than that from the water protons. Since the spatial location of the
MR signal in the frequency-encode direction is determined from the signal frequency, the fat
protons are registered at a slightly different position to that of the water protons, and so are
misregistered in the direction of the weaker end of the readout gradient. This results in an
image in which the signals from fat have a different location to those of water, producing
artifactual high and low signal bands at anatomic sites where watery and fatty tissues
(e.g., kidney and perinephric fat) are in contact (Fig. 13). The degree of chemical shift
CHAPTER 2: TECHNIQUES FOR MRI OF THE LIVER 13
misregistration increases with greater field strength and reduced sampling bandwidth.
The artifact may be considerably smaller with a high bandwidth technique but this is
achieved at the expense of SNR. However, chemical shift is effectively eliminated with all
types of fat suppression.
Because of their slightly different resonant frequencies, fat and water protons can be in- or
out-of-phase with each other at different echo times on GRE images. At 1.5T, fat and water
are in phase at TEs of 4.4 msecs and 8.8 msecs etc., and out of phase at 2.2 msecs, 6.6 msecs
etc. With spin echo imaging the signals from fat and water are always in phase. When fat
and water protons co-exist within a voxel their opposed phases cancel each other, producing
a loss of signal. This is manifest as an edge artifact at the fat/water interface on opposed-
phase (OP) GRE images (Fig. 14). The artifact is not a feature of spin-echo sequences or in-
phase GRE images and again it is effectively eliminated with fat suppression. With very large
or exophytic tumors, the fat/water cancellation artifact on opposed-phase T1w images is
often helpful in determining the organ of origin and involvement of adjacent structures
(Fig. 15). While the presence of an intact edge artifact between a mass and adjacent organs
reliably excludes invasion, a disrupted artifact is suggestive of involvement (Fig. 15).
FAT SUPPRESSION
Fat suppression strategies are strongly recommended for abdominal imaging. They
eliminate ghosting artifact from the high signal of fat on T2w FSE (Fig. 16) images and
increase image contrast which accentuates gadolinium enhancement. For 2D imaging the
authors routinely use a fat suppression technique which selectively excites water protons
using a binomial pulse sequence. Compared with the standard frequency-selective method
of fat suppression, this approach is more efficient and less sensitive to radiofrequency
inhomogeneities. Consequently, more slices can be obtained for a given acquisition time and
fat suppression is more homogeneous. A simple 1:1 binomial pulse sequence uses two short
RF pulses of equal amplitude. The first pulse tips all spins by 458 after which a gap (2.2 msec
at 1.5T) is left for fat and water protons to become out-of-phase with each other. The second
RF pulse then tips the water protons over by another 458 into the transverse plane but
because the fat protons are pointing in the opposite direction they are tipped back 458
towards the z-axis so only the water component generates signal (Fig. 17). Although the
2.2 msec gap and the use of two RF pulses results in a longer excitation time than a simple
non-selective pulse sequence, the standard spectral fat-suppression method has an even
longer excitation time since the duration of the fat-sat pulse is approximately 10 msec.
T2-WEIGHTED IMAGING
FSE sequences have replaced CSE for T2w imaging because they allow better resolution for a
given acquisition time, or can use a shorter acquisition time to allow breathhold imaging.
With FSE, multiple 1808 refocusing pulses generate a train of echoes after each 908 excitation
pulse. Compared with conventional SE, FSE sequences achieve a reduction in scan time
which is proportional to echo train length (the number of phase-encoding lines acquired per
TR). Contrast in FSE is manipulated by the ordering of K space. Although each echo is
acquired with a different TE, the central lines of K space (i.e., the lowest ordered phase
encoded steps), which determine image contrast are acquired at the time of the “effective”
TE (TEeff). Theoretically TEeff may be assigned to any echo in the echo train although the
degree of user flexibility varies between vendors. The low order phase encoding signals are
obtained when the phase encoding gradient has the smallest amplitude so there is only
minimal dephasing of the echo and signal is maximized. Traditionally, T2w images have
been regarded as the most useful for depicting pathology, but liver-to-lesion contrast
decreases as echo train length increases, and this has resulted in a rather poor detection rate
for solid lesions on T2w FSE images (Fig. 18). The lower liver/lesion contrast with FSE
is multi-factorial. Firstly, normal liver has a comparatively short T2 relaxation time of
14 PART A: WHY, WHEN, AND HOW TO PERFORM MRI OF THE LIVER
approximately 40 msec, probably due to hepatic iron stores and other paramagnetic
substances which induce signal loss due to diffusion in adjacent water molecules.
Metastases, hemangiomas and simple cysts have T2 relaxation times of approximately
80 msec, 160 msec and 240 msec respectively. Although cysts and hemangiomas are well
seen on FSE, the inherent contrast between metastases and normal liver is relatively poor.
Moreover FSE sequences are fairly insensitive to susceptibility so the signal intensity of the
normal liver is slightly higher than with conventional SE. Secondly, FSE is influenced by
magnetization transfer (MT) which causes solid tissues to lose signal intensity (tumor tissue
is particularly affected) so lesion/liver contrast is further reduced. Fluid is unaffected by MT,
so benign lesions are well shown on FSE (Fig. 19). Although MT occurs to some extent in all
multislice techniques, the 1808 off-resonance pulses in FSE significantly increase MT
compared with conventional SE sequences. Restricted protons bound to macromolecules
resonate over a much larger frequency range than free water. An RF pulse applied (on
resonance) to a particular imaging section will act as an off-resonance pulse to other imaging
sections in a multi-section acquisition and saturate the restricted pool. The saturation is
transferred to adjacent mobile water protons and a net loss of signal occurs. Solid lesions
have a large number of restricted protons and show large magnetization transfer effects
compared to normal liver (Fig. 20). In general as echo train length increases (i.e., with more
1808 refocusing pulses) MT effects increase, and lesion/liver contrast decreases even further.
Gradient echo sequences are less affected by MT. MT effects and therefore lesion/liver
contrast may be improved by reducing echo train length, but the longer acquisition time
then necessitates respiratory gating to minimize motion artifacts. FSE sequences are also
subject to blurring from tissues with short T2. Late echoes subject to pronounced T2 decay
have a lower signal amplitude than echoes acquired at the start of data acquisition. For this
reason it is important to minimize echo spacing as much as possible. Also the increased
signal intensity of fat on FSE (due to a phenomenon known as J-coupling) increases motion
artifact, so fat suppression is strongly recommended with FSE (Fig. 16). Effective echo times
of 80–100 msec will maximize image contrast for lesion detection.
GREATER T2 WEIGHTING
More heavily T2w images are necessary to distinguish non-solid benign and malignant
hepatic lesions. To avoid overlap between the signal intensity characteristics of solid
lesions (particularly hypervascular metastases) and hemangiomas on standard T2w FSE
images, TE must be increased to approximately 180 msec. At even longer echo times,
hemangiomas—which have a T2 relaxation time of approximately 160 msec—start to lose
their signal and mimic metastases (Fig. 21). Breathhold FSE sequences are suitable for
characterizing lesions because the strong MT effects which effect only solid tumors
accentuate the difference in the signal intensity between benign and malignant lesions.
With single shot FSE (SSFSE) techniques multiple echoes which fill the whole of K space are
acquired after a single excitation pulse. Consequently scan times are extremely short with
measurement times for a single image of approximately one-second. Since the excitation
pulse is not repeated, SSFSE sequences do not have a true repetition time i.e., TR is “infinite”.
The denoted TR here is the acquisition time of a single slice and this is directly influenced by
the number of phase encoding lines. Single shot FSE (SSFSE) sequences with partial Fourier
techniques (e.g., half Fourier acquisition single shot fast spin echo [HASTE]) are particularly
effective in distinguishing benign and malignant lesions. With the HASTE technique just
over half of K space is encoded with a single echo train and the remainder filled with half
Fourier transformation. Images are obtained in less than a second with virtually no motion
artifact even during free breathing so HASTE is particularly useful in uncooperative
patients. This is a single slice technique so MT effects are less significant, but because the
duration of the echo train (approximately 400 msecs) is considerably longer than the T2
relaxation time of solid liver lesions (approximately 80 msecs), there is relatively little signal
from short T2 tissues by the end of image acquisition. Conversely, tissues with a longer T2
CHAPTER 2: TECHNIQUES FOR MRI OF THE LIVER 15
exhibit a much higher signal and are well seen. Also, while blurring of short T2 tissues
maybe particularly pronounced on HASTE, tissues with a longer T2 are generally well-
defined. This results in improved sharpness and conspicuity of cysts and small fluid
collections (Fig. 22). Edge definition is enhanced by using a high acquisition matrix, and also
improves when the effective TE approaches the end of the echo train (long effective TE
sequence) so that the high spatial frequencies which determine edge definition are collected
at the earlier echoes. At our institution we use HASTE images at a TE of approximately
100 msecs to characterize focal liver lesions because motion artifacts are consistently
negligible regardless of patient co-operation and non-solid benign lesions have a crisp well-
defined edge (Fig. 23). However, HASTE is limited by a low SNR compared with standard
FSE sequences, such that small low-contrast lesions are not well shown. Heavily T2w SSFSE
(full Fourier) with TE of R500 msecs, and HASTE (half-Fourier) sequences are used to image
the fluid filled biliary tree for MR cholangiopancreatography. Thick-slab SSFSE with a slab
thickness of 30–80 mm and thin-slice multi-slice HASTE images have a complementary role.
Thick-slab images are obtained with an extremely long TE to provide complete suppression
of background tissue and an overview of bile duct anatomy, while thin-slice HASTE images
with a slice thickness of 2–4 mm are obtained at a more moderate TE of approximately
100 msec to improve the visualization of fine structures since small caliber structures may be
obscured with longer TE (Fig. 24).
STIR SEQUENCES
FSE-short tau inversion recovery (STIR) imaging is a valuable alternative to FSE for lesion
detection. Breathhold FSE STIR has been shown to be superior to CSE and non-breathhold
STIR for the detection of hepatic tumors. With STIR the inversion time is chosen to null
the signal from fat and image contrast is provided by the additive effects of T1 and T2.
Compared with a standard chemical shift saturation pulse, fat suppression on STIR is
generally more homogeneous; at 1.5T an inversion time of approximately 150 msecs is
optimum. Standard STIR is characterized by a fairly short TE (approximately 30 msecs) to
maximize both the SNR and the number of slices obtained with a given TR. With FSE STIR,
longer TEs are feasible without incurring time penalties, so allowing a greater T2 contribution
in the images, which should improve the conspicuity of abnormalities. The combination of fat
suppression and combined T1 and T2 effects with FSE STIR results in greater soft tissue
contrast and increased lesion conspicuity (Fig. 25). However, because this is a breathhold
sequence with a long echo train, MT effects are still a limitation and relatively fewer slices per
breathhold are possible due to the initial 1808 inversion pulse. One should also be aware that
the signal of other substances such as melanin or proteinacious fluid which have T1 relaxation
times similar to fat may also be suppressed on STIR. Breathhold STIR may also be used to
characterize lesions. As with other types of heavily T2-weighted imaging, non-solid benign
lesions are typically homogeneous and markedly hyperintense with crisply defined borders.
HYBRID SEQUENCES
Turbo gradient spin echo (TGSE) or gradient-recalled spin echo (GRASE) sequences combine
echo planar (EPI) and FSE techniques. All echoes are gradient refocused with EPI whereas
with FSE they are refocused spin echoes. With TGSE techniques, gradient echoes are
generated before and after each spin echo. Image contrast results from the spin echo
component of the sequence while the gradient echoes contribute only to resolution.
Compared with FSE, TGSE is more efficient, MT effects are less and the high signal intensity
of fat is less pronounced. However, image quality is variable due to degradation by motion
and susceptibility artifacts. The EPI component (the number of gradient echoes) of the
sequence is extremely sensitive to phase errors and variations in magnetic field homogeneity
and at the longer EPI factors required for breathhold imaging artifacts maybe severe
(Fig. 26). When TGSE is implemented with a shorter EPI factor, artifacts are reduced and
lesion to liver contrast is improved compared with FSE (Fig. 27). However longer acquisition
times prohibit breathholding and efficient respiratory gating is required to minimize
motion artifact.
16 PART A: WHY, WHEN, AND HOW TO PERFORM MRI OF THE LIVER
True fast imaging with steady state precession (True FISP, Siemens) or balanced fast field
echo (FFE, Philips) sequences have the highest signal of all steady state sequences. The
technique employs balanced gradients in all three axes at the time of radiofrequency
excitation which compensate for motion-induced phase shifts. Although the contrast here is
a function of T1 divided by T2, high contrast images which are essentially T2-weighted are
generated. As with any steady-state free precession technique, both transverse and
longitudinal magnetization contribute to the signal. True FISP is useful in patients who
are unable to breathhold because each slice is acquired in approximately one second. It
provides an excellent and quick anatomical survey, and is particularly useful for
demonstrating vessels and lymphadenopathy (Fig. 28). Limitations include a sensitivity to
magnetic field inhomogeneities which may result in fairly severe artifact, and low liver/
tumor contrast (Fig. 29). True FISP is an out-of-phase GRE sequence so it is fairly sensitive to
fatty infiltration. If severe, this produces an abnormally low liver signal on True FISP images,
and occasionally an edge artifact at the interface between the liver parenchyma and the
intra-hepatic vessels (Fig. 30). In milder cases of fat deposition, no signal change will be
discernible on True FISP images since the signal intensity of normal liver is already low.
T2-WEIGHTED GRE
Contrast-to-noise ratio in the liver is generally poor on long TE GRE images. However,
because T2*-weighted images are sensitive to susceptibility they are useful for depicting iron
deposition in siderotic nodules, haemochromatosis, and the breakdown products of
resolving hematomas (Fig. 31). They are also recommended for detecting focal lesions
following the injection of superparamagnetic iron oxide particles (SPIO) but parameters
must be optimized to minimize noise, to maximize the signal from solid lesions, and to
minimize motion-related artifact which may be greater at longer TEs (Fig. 32). A detailed
description of this pulse sequence can be found in chapter 3.
T1-WEIGHTED IMAGING
Contrary to early experience we now find that many lesions are better seen on T1w than on
T2w images (Fig. 33). Gradient echo sequences are routinely used for T1w imaging because
gradient refocusing of echoes is simpler and more efficient than radio frequency refocusing.
Shorter TEs and more slices for a given TR are achieved with gradient echo sequences than
with otherwise comparable SE sequences. With multi-section spoiled GRE sequences (e.g.,
FLASH, Siemens; SPGR, IGE; FFE, Philips) the entire liver is imaged in a single breathhold
with high SNR and minimum inter-slice spacing. For 2D imaging TRs of 100–200 msec
facilitate acquisition times short enough for breathholding with maximum SNR. TE should
be kept as short as possible to maximize the number of available slices and minimize T2
influence, and a flip angle of 70–908 is chosen to provide optimum T1 contrast (Fig. 34).
Providing fat and water co-exist within a voxel, opposed-phase T1w GRE images are very
sensitive to the fat content of tissues. This is particularly useful in liver imaging where the
combination of in-phase (IP) and opposed-phase (OP) GRE images (chemical shift imaging)
provides a reliable diagnosis of focal or diffuse fatty infiltration. The normal liver has the
same signal intensity on IP and OP GRE images and it is always greater than the signal
intensity of the spleen. Fatty and non-fatty liver have the same signal on IP images but on OP
images the fatty liver shows a reduced signal intensity which may be equal to or less than
that of the spleen (Fig. 35). We obtain IP and OP images in all patients undergoing liver MR
because they are sensitive to quite small degrees of fatty infiltration. Chemical shift imaging
CHAPTER 2: TECHNIQUES FOR MRI OF THE LIVER 17
also aids lesion characterization by demonstrating fat within tumors (see chapter. 7). In
patients with a fatty liver, lesions may be isointense with the lower signal of fatty liver on OP
images but lesion-to-liver contrast is greater on IP images (Fig. 36). Chemical shift imaging is
also able to distinguish signal loss due to fatty liver from signal loss caused by excess iron
deposition and to confirm the presence of surgical clips which may otherwise be wrongly
interpreted as pathology (Fig. 37). Signal loss due to metal is caused by susceptibility effects
which become more pronounced with increasing TE. In general in-phase TE is longer than
opposed-phase TE so signal loss due to iron excess or clips is more pronounced on IP T1.
IP and OP images may be acquired together as part of a multi-echo technique in which
two echoes per excitation pulse are acquired simultaneously. This approach avoids misreg-
istration and facilitates the exact replication of anatomical structures at each TE. However,
SNR may be reduced because the acquisition of two echoes in a breathhold necessitates
the use of a higher bandwidth compared with a single echo sequence (see also chapter 7).
Effective imaging with extracellular space gadolinium-based contrast agents requires rapid
sequential acquisitions at various critical time points after bolus injection. A T1-weighted
GRE sequence with the shortest possible TE should be used to maximize the number of slices
in order to obtain sufficient anatomic coverage in a breathhold. 2D spoiled GRE sequences
have been widely used and have an established record of success but they are limited by
relatively thick sections and inter-section gaps. Coverage of the entire liver in acquisition
times short enough for breathholding comes at the expense of reduced spatial resolution and
partial volume averaging which may limit the diagnosis of sub-cm lesions. The introduction
of fast 3D T1w GRE sequences has overcome many of the problems associated with 2D
sequences. 3D imaging methods give a higher SNR and thinner effective slice thickness than
2D methods, with no inter-slice gap or crosstalk. Combined with interpolation algorithms to
facilitate high resolution matrices and the repeated application of fat suppression pulses,
current 3D GRE sequences facilitate high resolution dynamic Gd-enhanced imaging with
improved contrast resolution and superior depiction of small tumors. Flexible parameters
allow scan times to be adjusted to accommodate the breathholding capacity of individual
patients. Additionally parameters may be manipulated to produce isotropic voxels for
optimal 3D reconstruction (Fig. 38). Isotropic resolution is achieved at the expense of
anatomic coverage but this is largely overcome by parallel imaging techniques. In most
patients a parallel imaging factor of two enables breathhold isotropic imaging of the whole
liver in approximately 20 seconds. 3D T1w GRE sequences are obtained with minimum TR
and TE to maximize the number of image sections for a given time. With short TEs, magnetic
susceptibility artifacts are less apparent because there is less time for signal loss to occur.
Minimum TEs also result in opposed-phase images which have a signal cancellation artifact
at fat/water interfaces that may reduce the conspicuity of small abnormalities, particularly
near the liver surface. The fat suppression pulse is an essential addition to this technique
because it eliminates the fat/water cancellation artifact and accentuates gadolinium
enhancement. A flip angle of approximately 158 minimizes the saturation of stationary
tissues for the simultaneous display of liver parenchyma and hepatic vessels (Fig. 39) while a
flip angle of 258 or more is used for dedicated MR angiography (Fig. 40). A more detailed
description of 3D T1w GRE sequences for dynamic Gd-enhanced imaging is provided in
chapter 3. Rapid 3D T1w GRE sequences may be used for any application requiring
gadolinium enhancement. They are particularly helpful in distressed patients who are
unable to stay still for higher resolution imaging with longer scan times (Fig. 41).
Magnetization-prepared GRE sequences generate T1w images in less than a second with a
reasonable SNR and minimum motion artifact even during free breathing. This approach is
extremely valuable in patients who have difficulty breathholding but is not generally
recommended for routine liver imaging because SNR is reduced and image contrast is less
reliable than with multi-slice spoiled GRE techniques. Also, since this is a single slice
technique, timing image acquisition to coincide with arrival of the gadolinium bolus in the
18 PART A: WHY, WHEN, AND HOW TO PERFORM MRI OF THE LIVER
tissue of interest is more difficult because timing will vary from slice to slice. The preparation
and acquisition of each individual slice takes approximately one second and is completed
before the next slice is acquired, so a stack of sequential slices is selected to encompass the
whole liver. Typically 128 lines of K space are measured per slice. Magnetization-prepared
GRE sequences comprise an initial 1808 inversion pulse followed by a delay [the inversion
time (TI)] and a standard spoiled GRE sequence which employs a small flip angle and a short
TR and TE. Image contrast is determined by the preparatory phase and depends primarily
on the effective TI, which is the time from application of the 1808 pulse to acquisition of the
central phase encoding lines. As with all inversion recovery sequences, different inversion
times can be selected to suppress the signal from specific tissues and vary T1 contrast. A non-
slice selective inversion pulse produces consistent image contrast across all slices providing
the TR is long enough to allow sufficient recovery of the longitudinal magnetization between
each inversion pulse (Fig. 42).
In patients who are unable to perform repeated breathholds, we obtain sequential True
FISP and HASTE images during free breathing with navigator pulses applied to the HASTE
sequence to eliminate mis-registration. A stack of magnetization-prepared GRE T1w images
are then obtained before and at the arterial, portal and equilibrium phases after a bolus
injection of gadolinium (Fig. 43). More delayed T1w images are obtained in selective cases.
We use a low-resolution version of the same sequence for test bolusing with gadolinium. For
more details of the test bolus technique see chapter 3.
ILLUSTRATIVE FIGURES
REFERENCES
1. Mitchell DG. MRI Principles. Philadelphia: W.B. Saunders, 1999.
2. Elster AD. Questions and answers in Magnetic Resonance Imaging. St. Louis, Missouri: Mosby, 1994.
3. Morrin MM, Rofsky NM. Techniques for liver MR imaging. Magn Reson Imaging Clin N Am 2001;
9:675–696.
4. Lee VS, Lavelle MT, Krinksy GA, Rofsy NM. Volumetric MR imaging of the liver and applications. Magn
Reson Imaging Clin N Am 2001; 9:697–716.
5. Regan F. Clinical applications of half-fourier (HASTE) MR sequences in abdominal imaging. Magn Reson
Imaging Clin N Am 1999; 7:275–288.
6. Barish MA, Jara H. Motion artifact control in body MR imaging. Magn Reson Imaing Clin N Am 1999;
7:289–302.
7. Blaimer M, Breuer F, Mueller M, Heidmann RM, Griswold MA, Jakob PM. SMASH, SENSE, PILS,
GRAPPA: how to choose the optimal method. Top Magn Reson Imaging 2004; 15:223–236.
8. Margolis DJ, Bammer R, Chow LC. Parallel imaging of the abdomen. Top Magn Reson Imaging 2004;
15:197–206.
CHAPTER 2: TECHNIQUES FOR MRI OF THE LIVER 19
Figure 1. Image quality: Breathhold T1w and T2w images in a patient with multiple metastases illustrating
the typical image quality of current MR pulse sequences. Sub-second HASTE (A) and 3D T1w GRE (VIBE)
images before (B) and after (C) gadolinium showed multiple lesions of varying size with high lesion/liver
contrast and negligible motion artifact.
rf rf
a+
Gr a+
Gr
a+ a+
Signal Signal
Short TE Long TE
(A) (B)
Figure 2. Influence of receiver bandwidth on signal to noise: The readout gradient may be
applied at a higher or lower sampling frequency (high or low bandwidth sequences). With a high
bandwidth, the readout gradient (shown above the line) has twice the area (aD) of the
dephasing gradient (shown below the line) and the peak of the echo occurs when the amount of
rephasing equals dephasing (i.e., when the areas aD are equal). With a lower bandwidth, there
is less noise in the image because fewer frequencies are sampled but it takes longer to sample
the echo, so echo time is increased (B). In this example, as sampling bandwidth is increased
from 200 Hz/pixel on (C) to 380 Hz/pixel on (D) SNR decreases due to more noise in the higher
bandwidth image.
rf rf
a+ b+
Gr Gr
a+ b+
Figure 3. Asymmetric echo sampling used
to shorten TE: When the echo is sampled
Signal Signal asymmetrically (B), only part of the echo is
sampled. Compared with symmetric sampling
TE TE of the whole echo (A), the peak of the echo
(A) (B) occurs earlier and so TE is shorter.
20 PART A: WHY, WHEN, AND HOW TO PERFORM MRI OF THE LIVER
Figure 6. The effect of respiratory gating on motion artifact reduction: Respiratory gated T2w
FSE in two patients with regular (A) and irregular (B) breathing patterns. Motion artifact was
effectively suppressed on (A) but “ghosting” is considerable on (B). In a third patient also with an
erratic breathing pattern, breathhold T2w FSE (C) showed much better image quality than
respiratory gated T2w FSE (D) images.
CHAPTER 2: TECHNIQUES FOR MRI OF THE LIVER 21
Figure 12. Effect of pulse sequence and echo time on severity of susceptibility artifact: The
signal void (arrows) due to surgical clips at the right hepatectomy resection margin is less
severe on OPT1w (TE 2.4 msec) (A) than on IPT1w imaging (TE 4.7 msec) (B). The artifact
appears larger and more pronounced on (B) due to greater dephasing at the longer TE. The
artifact is imperceptible on HASTE (C) due to the multiple 1808 refocusing pulses, which
minimize susceptibility. Surgical clips were confirmed on the corresponding CT image (D).
Figure 13. First-order chemical shift mis-registration artifact on low bandwidth T2w GRE
image: Relative to the signal from fat, water from the kidney is incorrectly mapped along the
frequency encoding direction toward the right. This is illustrated by a black signal void (arrow)
at the lateral margin of the left kidney and a high signal band (arrow) on the medial aspect.
CHAPTER 2: TECHNIQUES FOR MRI OF THE LIVER 23
y y y
x x x
(A) (B) (C)
Figure 17. Fat suppression using a 1:1 binomial water excitation pulse sequence: After the first 458 pulse
which tips the equilibrium magnetization toward the transverse plane, (A) a delay of 2.2 msec allows the fat
and water protons to precess out of phase with each other and point in opposite directions. (B) When a
second 458 pulse is applied, (C) the fat protons (light grey) are tipped back into the z-axis; only the water
protons (dark grey) are tipped into the transverse plane so signal intensity in the resulting image is mostly from
water.
24 PART A: WHY, WHEN, AND HOW TO PERFORM MRI OF THE LIVER
Figure 18. Reduced liver/lesion contrast on breathhold T2w FSE in two patients with advanced cirrhosis
and hepatocellular carcinoma: A 6 cm HCC (arrow), which was highly conspicuous on SPIO-enhanced
T2w GRE (A), was barely visible on the corresponding unenhanced T2w FSE image (B). In a second patient a
poorly seen HCC (arrow) on unenhanced T2w FSE (C) was more conspicuous on the corresponding T2w FSE
images after SPIO (D) and seen even more clearly on post-SPIO T2w GRE imaging (E).
Figure 19. Benign versus malignant lesion conspicuity on T2w FSE: Breathhold T2w
FSE images showed a highly conspicuous high signal intensity cyst (short arrow), and a
metastasis (long arrow), which was less conspicuous due to MT effects and its shorter T2
relaxation time.
Figure 20. Effect of magnetization transfer on the conspicuity of solid lesions: In two patients with
colorectal metastases, breathhold T2w FSE images were acquired with identical parameters but as a multi-
section acquisition in (A) and (C) and with single slices in (B) and (D). Lesion/liver conspicuity was less on (A)
than (B) because of MT effects. Tissue contrast is diminished in multi-slice acquisitions due to partial
saturation of macromolecules within the acquisition volume by radiofrequency pulses applied to a single
image section within the volume. The central necrotic components of the posterior metastasis in both (C) and
(D) (arrow) were unaffected by MT and had a high signal intensity relative to the background liver on both
images. The solid component of the lesion was highly conspicuous on the single slice acquisition but difficult
to see on the multislice acquisition. (Continued)
CHAPTER 2: TECHNIQUES FOR MRI OF THE LIVER 25
Figure 21. Effect of increasing TE on the signal intensity of hemangiomas: Breathhold T2w FSE images
were obtained at TEs of 90 (A), 180 (B), and 320 msec (C). The hemangioma had a signal intensity similar
to CSF on (A) and (B), but at the longest TE (C), the lesion had less signal intensity than CSF due to its
shorter T2 relaxation time.
Figure 22. Effectiveness of HASTE in depicting small cysts: Simple cysts are well-
defined and have a very high signal intensity on HASTE images. In a patient with co-
existent cysts and metastases, a small cyst was highly hyperintense and sharply
defined (arrow) while the metastasis was less well-defined and relatively inconspicuous
due to a shorter relaxation time and blurring of short T2 tissues.
Figure 23. HASTE verses heavily T2w breathhold FSE for depicting small cysts: On heavily T2w FSE
images (TE 180 msec) (A), although the large right lobe cyst had a very high signal intensity and was well-
defined, the small cyst in the left lobe (arrow) was ill-defined and difficult to characterize. Ghosting artifacts
from the aorta and large cyst were pronounced despite patient compliance. On HASTE imaging (B) the small
cyst was conspicuous, sharply defined and more reliably characterized. Motion artifacts were also negligible
on HASTE images.
26 PART A: WHY, WHEN, AND HOW TO PERFORM MRI OF THE LIVER
Figure 24. Complementary role of thick slab MRC and thin-slice HASTE: Coronal/oblique thick slab
SSFSE imaging (A) showed a tortuous and dilated biliary tree and questionable filling defects in the CBD.
Axial thin-slice HASTE images (B and C) clearly showed multiple, dependent filling defects in the CBD and
intra-hepatic ducts and in the gallbladder. Coronal/oblique thick slab SSFSE image (D) in a different patient
with Caroli’s disease showed an apparently solitary filling defect in the distal CBD (arrow). Axial thin-slice
HASTE image (E) showed additional stones in the right-sided intra-hepatic ducts (arrow).
Figure 26. Artifact associated with long EPI factor on breathhold TGSE: There was
high lesion/liver contrast on the breathhold TGSE images, but they were degraded by
phase artifact (ghosting) due to the long gradient echo train (EPI factor).
CHAPTER 2: TECHNIQUES FOR MRI OF THE LIVER 27
Figure 27. Lesion/liver contrast on non-breathhold TGSE compared with T2w FSE: In a patient with a
regular breathing pattern, motion artifacts were effectively eliminated with respiratory gating, and lesion/liver
contrast was better on non-breathhold TGSE (A) than on non-breathhold FSE images (B). In a second patient
with more erratic breathing, respiratory gating was ineffective and non-breathhold TGSE images (C) were
severely degraded by motion.
Figure 28. Role of True FISP imaging: Single breathhold coronal (A) and axial (B) True FISP
images illustrate the value of True FISP in obtaining a rapid anatomical survey of the abdomen
with high contrast and negligible motion artifact. Axial True FISP images in two other patients (C
and D) showed conspicuous nodal disease (arrows). Axial True FISP image in a fourth patient
with non-occluding thrombus of the portal vein showed high contrast between the low signal
intensity thrombus (arrow) and the high signal of flowing blood (arrow) (E). A coronal/oblique
thick-slab MRCP image (F) in a fifth patient showed a typical hepatic artery compression
artifact at the bile duct bifurcation (arrow) mimicking a filling defect. The corresponding
coronal/oblique True FISP image (G) demonstrated the hepatic artery (arrow) and confirmed its
position adjacent to the signal void.
28 PART A: WHY, WHEN, AND HOW TO PERFORM MRI OF THE LIVER
Figure 30. Signal intensity change on True FISP due to fatty infiltration: True FISP image in a
patient with a fat-containing adenoma (arrow) (A) illustrated a lower signal intensity in the lesion
than in normal liver tissue. In a second patient with fairly severe fatty change as illustrated by a
marked drop of liver signal on OPT1w (B) compared with IPT1 images (C), the whole liver had an
abnormally low signal on the corresponding True FISP image (D). Note that the normal liver in (A)
had a signal intensity close to the signal intensity of spleen, whereas the signal intensity of the
fatty liver in (D) was much lower than the spleen.
Figure 31. Susceptibility effects on T2w GRE: The normal liver has a signal intensity only slightly lower
than adipose tissue on unenhanced T2w GRE images (A) and contrast is poor. In a patient with excess
iron deposition in the liver and spleen due to transfusional siderosis, there was a marked reduction in the
signal intensity of the liver and spleen on unenhanced T2w GRE images (B). In a different patient,
siderotic nodules were clearly seen as punctate areas of very low signal on T2W GRE (TE 14 msec)
images (C). They were much less conspicuous on the corresponding in-phase T1W GRE (TE 4.7 msec)
image (D) and undetectable on the opposed-phase T1w GRE (TE 2.4 msec) image (E). Susceptibility
effects diminish as TE is reduced. (Continued)
CHAPTER 2: TECHNIQUES FOR MRI OF THE LIVER 29
Figure 32. Optimized SPIO-enhanced T2w GRE imaging: T2w GRE image obtained with optimized
parameters after SPIO (A) illustrates high lesion-to-liver contrast and negligible artifact. Note the improved
conspicuity of sub-cm lesions (arrows) compared with corresponding unenhanced IPT1w GRE (B) and
HASTE (C) images. In a second patient with a large colorectal metastasis and numerous satellite lesions the
individual tumor deposits were more clearly defined on SPIO-enhanced T2w GRE (D) than on unenhanced
IPT1 GRE (E) and HASTE (F) images.
Figure 33. Comparison of lesion/liver contrast on current breathhold, non-contrast T1w and T2w
sequences: T2w FSE (A), HASTE (B), STIR (C), True FISP (D), OPT1w GRE (E), and IPT1w GRE (F)
images were obtained in a patient with multiple colorectal metastases. Image quality was good and motion
artifacts were negligible on all images. Lesion/liver contrast was poor on (A) probably due to MT effects.
Compared with (A), lesion/liver contrast was better on (B) probably due to less MT, and better on
(C) because of the additive effects of T1 and T2 contrast. Only the central necrotic portion of the larger lesion
was visible on (D). Compared with the four T2w sequences, lesion/liver contrast was considerably better on
both T1w sequences (E and F), due to greater SNR and lack of MT effects. (Continued)
30 PART A: WHY, WHEN, AND HOW TO PERFORM MRI OF THE LIVER
Figure 34. Influence of flip angle on T1 contrast: IPT1w images with 258 flip angle (A), 458
flip angle (B), 658 flip angle (C), 858 flip angle (D) were obtained on a normal subject; contrast
between the liver and spleen and the liver and the intra-hepatic vasculature increases with
increasing flip angle and is optimum at a flip angle of 858.
Figure 35. Fatty infiltration of the liver identified by T1w GRE in-phase (TE 4.7 msec) and
opposed-phase (TE 2.4 msec) (imaging): the normal liver has the same signal on IPT1w (A)
and OPT1w images (B) and liver signal is brighter than the spleen in both. In a different patient
with diffuse fatty infiltration, the liver had a normal signal on IPT1w (C) but a much lower signal on
OPT1w images (D) and on (D) the liver signal was lower than the signal intensity of the spleen. An
area of focal fatty sparing (arrow) had a higher signal than the liver and spleen on (D) and was
homogeneously isointense with the adjacent liver on (C).
CHAPTER 2: TECHNIQUES FOR MRI OF THE LIVER 31
Figure 37. Signal loss due to susceptibility effects confirmed with IP and OP T1w GRE
imaging: In a patient with hemochromatosis the signal intensity of the liver was slightly lower
than the signal intensity of the spleen on OPT1w images (TE 2.4 msec) (A). On the corresponding
IPT1w image (TE 4.7 msec) (B), there was further signal loss and the liver signal was markedly
lower than spleen due to greater susceptibility at the longer TE. In a different patient a signal void
due to a small surgical clip at the site of a previous cholecystectomy (arrow) appeared larger on
IPT1w (C) than on OPT1w images (D).
Figure 38. The value of isotropic voxels for optimal reformatted images: Post-Gd portal
phase axial 3D T1w GRE source images were obtained with near isotropic voxels (A).
Reformatted images in coronal (B) coronal/oblique (C) and sagittal (D) planes, derived from the
same source data as (A), have comparable image quality.
32 PART A: WHY, WHEN, AND HOW TO PERFORM MRI OF THE LIVER
Figure 39. Vascular MIP images derived from a 3D T1w GRE acquisition obtained with a flip angle of 158 for
optimal parenchymal evaluation: Post-Gd arterial (A) and portal (B) phase MIP images show normal arterial
and portal vasculature. Arterial phase coronal/oblique source image (C) clearly shows a hypervascular tumor.
Figure 40. Comparison of parenchymal and vascular depiction on 3D T1w GRE images at flip
angles of 158 and 258: In a liver transplant patient with post-operative variceal bleeding, the
liver parenchyma was optimally depicted on 3D T1w GRE images obtained with a 158 flip angle
(A) and a small hematoma at the posterior surface of the right lobe was well shown (arrow).
Arterial (B) and portal (C) phase MIP images derived from the same dynamic acquisition showed
a patent arterial conduit, patent portal veins and large varices. In a second patient also with an
arterial conduit following liver transplantation, complete occlusion of the hepatic artery was
shown on the arterial phase MIP image (D) which was obtained from source data (E) acquired
with a 258 flip angle. On the subsequent portal phase MIP (F) the portal vasculature was well
depicted but there is poor soft tissue contrast on the corresponding source image (G).
Parenchymal detail is better with the lower flip angle.
CHAPTER 2: TECHNIQUES FOR MRI OF THE LIVER 33
Figure 41. Role of Gd-enhanced rapid 3D T1w GRE imaging for non-hepatic applications: In a patient with severe lower back pain following an
extended right hepatectomy for a large HCC (A), routine spinal imaging showed an extensive abnormality at L3/4 (B and C) but the images were
badly degraded by movement artifact due to the acquisition time of several minutes. Subsequent rapid 3D T1w GRE images obtained in axial (D)
and coronal (E) planes after bolus injection of gadolinium, each with an acquisition time of 20 seconds, provided excellent demonstration of the
abnormality with negligible motion artifact. These post-contrast images showed a large psoas abscess communicating with the L3/4 disc space,
and extending anteriorly to erode the posterior wall of a large infra-renal aortic aneurysm.
Figure 42. Image contrast on magnetization-prepared GRE T1w images using a non-selective
inversion pulse: There is minimal contrast change between sequential images (A–D) at a TR of
1200 msec due to adequate recovery of the longitudinal magnetization between each inversion
pulse. Since each slice is acquired in just over one second, the images which were obtained
during free breathing were free of motion artifact.
34 PART A: WHY, WHEN, AND HOW TO PERFORM MRI OF THE LIVER
Figure 43. Non-breathhold imaging in compromised patients: True FISP (A), HASTE (B) and
magnetization-prepared post-Gd T1w GRE images were obtained at arterial (C), portal (D), and delayed (E)
phases in an elderly non-compliant patient. A left lobe lesion with typical MR findings of focal nodular
hyperplasia was shown. The lesion was isointense with background liver on (A) and (B) but the hyperintense
scar was visible (arrow). The lesion displayed marked enhancement on (C), which faded rapidly on (D), and
there was delayed enhancement of the central scar on (E). In a second encephalopathic patient with end-
stage liver disease, HASTE (F) imaging showed an enlarged left lobe with increased signal intensity
consistent with edema. An infiltrative HCC occupying most of the small right lobe was well seen on immediate
post-Gd magnetization-prepared T1w GRE images (G). Liver/lesion contrast was less in the subsequent
portal phase (H). The corresponding True FISP image (I) showed a large volume of ascites but no abnormality
in the liver parenchyma. High quality motion-free images were obtained during free breathing in both patients.
CHAPTER 3
- Hepatocyte Agents . . . . 38
Gadobenate . . . . 39
Gadoxetic Acid . . . . 39
Mangafodipir . . . . 39
- Illustrative Figures . . . . 43
- References . . . . 43
35
36 PART A: WHY, WHEN, AND HOW TO PERFORM MRI OF THE LIVER
espite the inherently high contrast resolution of MR the use of contrast agents has
D great diagnostic value for liver imaging. In many cases lesions become much more
conspicuous, benign, and malignant masses are more reliably distinguished, and
contrast enhancement characteristics often allow a specific diagnosis of pathology.
Moreover, since the replacement of non-breathhold T2w sequences with breathhold T2w
sequences—which are less sensitive for detecting solid lesions—the role of contrast
enhancement has become even more important.
Several different contrast agents are available for liver MR (Table 1). Dynamic
gadolinium (Gd)-enhanced techniques have been widely used in liver MR for several years
and more recently liver-specific agents have been developed to increase and prolong the
contrast between normal and abnormal tissues and further aid lesion detection. These agents
target either the hepatocytes or Kupffer cells to allow an assessment of pathology at cellular
level so they also have particular value for characterizing hepatocellular lesions.
ECF Gd is administered as a rapid bolus at a dose of approximately 0.1 mmol/kg and the
first acquisition is timed to coincide with arrival of the contrast in the tissue or vessel of
interest during acquisition of the central lines of k-space, which determines tissue contrast
within the image. Best results are obtained when the contrast is administered by a power
injector at a rate of 2–4 ml per second followed by a 20–30 ml saline flush.
To enable sequential imaging of the entire liver during suspended respiration (w 20 secs), a
T1w gradient echo sequence with the shortest possible TE should be used to maximize the
number of slices for a given TR. 3D sequences are preferred because they result in a higher
signal-to-noise and thinner effective slice thickness than do 2D methods, with no inter-slice
gap or crosstalk. We recommend a 3D T1w GRE sequence which comfortably allows thin-
section coverage of the whole liver in a single breathhold. This technique known as VIBE,
(Volume Interpolated Breathhold Examination) was first developed for abdominal imaging
Extracellular Gd chelates (gadopentetate, etc.) Intravascular and extracellular fluid Positive T1 enhancement
Extra/intracellular T1 agents (gadobenate, Intravascular and extracellular fluid, then Positive T1 enhancement
gadoxetic acid) hepatocytes, biliary tract
Intracellular T1 agents (mangafodipir) Hepatocytes, then biliary tract Positive T1 enhancement
Superparamagnetic iron oxide agents Intravascular fluid initially, then Positive T1, then negative
(ferucarbotran, ferumoxides) reticuloendothelial cells of liver and spleen T1 and T2* enhancement
CHAPTER 3: CONTRAST AGENTS FOR LIVER MRI 37
on a Siemens system by Rofsky and colleagues in 1998 but other manufacturers now offer
comparable sequences (e.g., Philips WAVE, GE FAME). The use of a short repetition time
(TR) and echo time (TE) facilitates a higher resolution matrix and the intermittent
application of a chemically-selective fat saturation pulse before each partition loop achieves
homogeneous fat suppression with minimal time penalty. The parameters we use are as
follows: TR 3.8 msec, TE 1.55 msec, flip angle (FA) 158 and bandwidth (BW) 490 Hz/pixel.
Acquisition of k-space is linear. We use a slightly asymmetric echo in the read out direction
(kx), to reduce TE and TR although at the expense of some signal-to-noise. With symmetric
echoes, the peak of the echo and the acquisition of the central lines of k-space occur at the
center (echo position 50%) of echo sampling. With asymmetric sampling the central parts of
k-space are acquired earlier at an echo position of 38%; approximately 80% of k-space is
filled and the remainder encoded by zero-filling. The matrix in the readout direction is 256.
We do not use partial Fourier in the in-plane phase encoding direction (ky) and acquire
approximately 125 phase encoding points. We use a rectangular (typically 75%) field of
view (FOV) of 280–400 mm depending on patient size to yield a pixel size of about 1.6 mm
in the frequency encoded direction and 2.4 mm in the phase-encoding direction. Partial
Fourier reconstruction is applied in the slice-select direction (kz) with 48–53 data points
interpolated to achieve 72–80 partitions. A slab thickness of 180–240 mm provides full
coverage of the liver with an interpolated slice thickness of 2.5–3.0 mm. The small FA of 158
minimizes saturation to maximize the signal from stationary tissues for simultaneous
display of liver parenchyma and hepatic vessels (Fig. 2). Fat suppression is particularly
helpful because it increases the dynamic range within the image and accentuates Gd
enhancement (Fig. 3).
Compared with 2DFT techniques at a comparable voxel size, 3DFT techniques theoretically
generate images with higher signal-to-noise ratio, so even when parameters are
compromised the image quality of 3D imaging is usually superior to that of conventional
2D imaging. VIBE is a very flexible sequence and can be easily adapted to achieve full liver
coverage in acquisition times as short as 12 seconds. In patients with a reduced
breathholding capacity we increase the effective slice thickness to 4 mm thereby reducing
the slab thickness needed to encompass the liver and we reduce the number of data points in
the phase encoding direction. Conversely if less anatomic coverage is required, the use of a
targeted coronal/oblique slab permits higher resolution images with an effective slice
thickness of 1–1.5 mm also in a breathhold period (Fig. 4). Furthermore, artifacts from
surgical clips, stents, and coils are minimal because short TE sequences are relatively
insensitive to susceptibility.
It should be recognized that a TE of approximately 2 msec produces opposed-phase
images which in the presence of fatty infiltration results in reduced liver signal and rarely
this may be decreased still further following Gd due to paradoxical enhancement. We no
longer use dedicated MRA sequences to image the hepatic vasculature, since the VIBE
sequence produces MIP reconstructions of comparable quality with excellent soft tissue
contrast (Fig. 5).
At least three phases of enhancement (arterial, portal, and equilibrium) are necessary. Of
these, the arterial-dominant phase—which is identified by marked enhancement of the
hepatic arteries, pancreas, and spleen with early filling of the main portal vein branches but
no enhancement of the hepatic veins—is the most crucial. The time window between the
arrival of contrast in the hepatic artery to filling of the hepatic veins is fairly narrow, and
many hypervascular lesions are only visible during this time (Fig. 6). In most patients,
effective arterial phase enhancement is achieved when the central lines of k-space are
acquired about 10–15 seconds after the end of injection. However, transit times from the
ante-cubital fossa to the hepatic artery have been shown to range from 8–32 seconds and an
38 PART A: WHY, WHEN, AND HOW TO PERFORM MRI OF THE LIVER
initial test dose of contrast with real time detection is recommended for optimal arterial
phase imaging.
We use a 2 ml test bolus of Gd-DTPA followed by a minimum 20 ml saline flush
injected at a rate of 4 ml/sec. A magnetization-prepared spoiled GRE T1w sequence
(Turbo FLASH, TR 1000 msec TE 1.58 msec, inversion time 500 msec, FA 808, BW 380 Hz/
pixel, matrix 123!256, slice thickness 10 mm) is positioned over the abdominal aorta at the
level of the diaphragm and a series of transverse images obtained at 1 second intervals.
Using a user-defined region of interest positioned over the aorta, an enhancement curve is
generated to determine peak enhancement. Based on an injection rate of 4 ml/sec the
optimum delay between the start of injection and start of data acquisition is calculated as—
the time to peak aortic enhancement plus the duration of Gd injection minus the interval
between starting acquisition and acquisition of the central lines of “k” space.
Lesions which have a predominantly arterial blood supply are best seen at the arterial
phase after Gd because the normal liver shows relatively little enhancement in this phase
(Fig. 6). Since most of the hepatic blood inflow (about 80%) arrives via the portal vein,
parenchymal liver enhancement is maximum at the portal phase which occurs 20–30
seconds later. Hypovascular lesions are most conspicuous in the portal phase, but rim
enhancement in the periphery of hypovascular tumors, which allows them to be
distinguished from benign lesions, is seen best on arterial phase images, so both phases
should be acquired in all patients (Fig. 7). The equilibrium phase occurs 2–3 minutes after
injection. At this time contrast between normal, and abnormal tissues is poor and some
lesions are no longer visible, but this phase is often helpful for characterization. When
lesions with a large fibrous component such as cholangiocarcinomas or hemangiomas are
suspected, further images should be obtained approximately 10 minutes after injection. At
this stage such lesions are often more conspicuous due to gradual accumulation of Gd
producing increasing enhancement (Fig. 8). For similar reasons, peritoneal tumor deposits
and extra-hepatic inflammatory change are also best demonstrated on delayed FST1w
images after Gd (Fig. 9).
Two types of specific contrast media have been developed for liver MR—agents based on Gd
or manganese chelates which shorten T1 relaxation time and target the hepatocytes, and
superparamagnetic iron oxide (SPIO) particles which cause a marked shortening of T2* and
are extracted by the reticulo-endothelial cells of the liver, spleen, and bone marrow. Non-
hepatocellular tumors become more conspicuous after contrast enhancement because
malignant lesions lack functioning Kupffer cells or hepatocytes, so their signal intensity is
unchanged while the surrounding liver becomes either hyperintense (T1 agents) or
hypointense (SPIO) (Fig. 10). Liver-specific contrast agents also help to distinguish between
benign and malignant hepatocellular lesions because they are taken up by functioning
hepatocytes and Kupffer cells in focal nodular hyperplasia (FNH), many liver cell adenomas,
and in the regenerating nodules of cirrhosis (Fig. 11). Well-differentiated hepatocellular
carcinoma (HCC) may also show contrast uptake but to a lesser extent.
HEPATOCYTE AGENTS
Three hepatocyte agents (gadobenate, gadoxetic acid, and mangafodipir) are currently
available. All three agents are paramagnetic, producing T1 shortening and positive
enhancement on T1-weighted images. They all have a greater relaxivity (the ability of a
contrast agent to selectively increase the relaxation time of water protons in tissue) than
the non-specific ECF agents. The relaxivity of contrast agents varies according to whether it
is measured in saline, plasma, or in tissue. In the liver the estimated T1 relaxivities of ECF
CHAPTER 3: CONTRAST AGENTS FOR LIVER MRI 39
Gd, gadobenate, gadoxetic acid, and mangafodipir are respectively 6.7, 30, 16.6, and
21.7 mmolK1, secK1. T1w GRE imaging is recommended for all three agents. To improve
the detection of sub-centimeter lesions the same T1w 3D VIBE sequence as is used with
Gd-DTPA is recommended.
Gadobenate
Gadoxetic Acid
Mangafodipir
Two SPIO agents are available for liver MR, ferumoxides (AMI-25, Endorem, Guerbet) and
ferucarbotran (SHU 555A, Resovist, Schering). The distribution of SPIO varies with particle
size. In the size range of 30–200 nm approximately 80% of the injected dose is taken up by the
normal liver. When SPIO particles are clustered within the macrophages, they induce local
field inhomogeneities which lead to rapid spin dephasing and loss of signal in normal liver
tissue on both T1w and T2w images. The effect is more pronounced on T2w images due to
Another random document with
no related content on Scribd:
"Silloin tajuat", jatkoi rouva Chick puristaen hiljaa neiti Toxin kättä
kiitokseksi hänen ystävällisestä huomautuksestaan, "että kaikki suru
on hyödytöntä ja että meidän velvollisuutemme on alistua".
"Ei, ei, ei", sanoi rouva Chick. "Kuinka voitte niin sanoa?"
"No niin, lapsi", sanoi rouva Chick, "sinä saat jäädä. Mutta se
minun on sanottava, että se on kummallinen valinta. Sinä oletkin
aina ollut kummallinen. Kuka toinen sinun ikäisesi tahansa olisi
luullakseni mielellään lähtenyt pois kaiken senjälkeen, mitä on
tapahtunut — rakas neiti Tox, nyt olen taas kadottanut nenäliinani."
Miksi kääntyivät tytön tummat silmät niin usein työstä sinne, missä
punaposkiset lapset asuivat? He eivät muistuttaneet hänelle
välittömästi äskeistä menetystä, sillä he olivat kaikki tyttöjä, neljä
pientä sisarusta. Mutta he olivat äidittömiä samoin kuin hän — ja
heillä oli isä.
Tuo talo oli ollut tyhjänä pitkän aikaa, monta vuotta. Vihdoin,
hänen ollessaan kotoa poissa, oli sinne muuttanut väkeä. Talo oli
korjattu ja uudelleen maalattu, ja akkunoiden kohdalla oli lintuja ja
kukkia, joten se näytti toisenlaiselta kuin ennen. Mutta Florence ei
koskaan ajatellut taloa. Lapset ja heidän isänsä olivat kaikki
kaikessa.
Kun ei kukaan muu talon väestä enää liikkunut ja kaikki valot oli
sammutettu, lähti hän hiljaa omasta huoneestaan ja sipsutti
äänettömin askelin portaita myöten alakertaan isänsä ovelle. Sitä
vasten hän tuskin hengittäen painoi kasvonsa ja päänsä ja huulensa
rakkautensa kaihossa. Joka ilta hän kyyristyi sen edessä kylmälle
kivilattialle kuunnellakseen edes isänsä hengitystä. Koko
sydämestään hän toivoi voivansa osoittaa isälleen rakkauttaan,
lohduttaa häntä, saada hänet sietämään yksinäisen lapsensa
hellyyttä. Jos hän olisi uskaltanut, olisi hän polvistunut hänen
jalkojensa juureen rukoilemaan sitä nöyrästi.
Kukaan ei sitä tiennyt tai ajatellut. Ovi oli aina lukossa, Dombey
sulkeutuneena sen taakse. Kerran pari hän lähti sieltä ulos, ja
talonväki sanoi hänen hyvin pian matkustavan maalle, mutta hän
asui yhä yksin huoneessaan, ollenkaan tapaamatta tai kyselemättä
Florencea. Ehkä hän ei edes tiennyt tyttärensä olevankaan samassa
talossa.
Vaistomaisesti Susan Nipper teki nyt samoin. Sillä välin Toots, joka
oli tullut hänen jäljessään yläkertaan tietämättömänä
herättämästään vaikutuksesta, ilmoitti saapumisensa koputtamalla
ovelle ja astui ripeästi sisään.
"Hyvää päivää, neiti Dombey. Minä voin oikein hyvin, kiitos. Mitä
teille kuuluu?"
"En oikein tiedä. Ei, en juuri vielä", sanoi Toots, istuutuen taas.
"Asia on niin — tarkoitan, neiti Dombey!"
Mutta vaikka Diogenes oli niin hullunkurinen koira kuin suinkin voi
tavata kesäisenä päivänä, kömpelö, ruma ja isopäinen, jota alinomaa
vaivasi se harhakuvitelma, että jossakin lähellä oli vihollinen, jolle piti
haukkua, ja vaikka se oli äreä ja kaikkea muuta kuin älykäs ja vaikka
sen silmät olivat kokonaan karvan peitossa, kuono hullunkurisen
näköinen, häntä ihan muodoton ja ääni karhea, oli se Florencelle
rakkaampi kuin kaikkein arvokkain ja kaunein koira, koska se oli
muistona Paulista, joka oli pyytänyt pitämään siitä huolta. Niin rakas
tämä ruma Diogenes tosiaankin oli hänestä ja niin tervetullut, että
hän tarttui Tootsin jalokivillä koristettuun käteen ja suuteli sitä
kiitollisena. Ja kun Diogenes vapaaksi päästyään syöksyi ylös portaita
ja ryntäsi sisään (kuinka vaikeata ensin olikaan saada se ulos
vaunuista!), pujahtaen kaikkien huonekalujen alle ja kiertäen
kaulassaan riippuvan pitkän ketjun tuolien ja pöytien jalkojen
ympärille ja sitten tempoen sitä, kunnes sen silmät tulivat
luonnottoman selvästi näkyviin, melkein pullistuessaan ulos
kuopistaan, ja kun se murisi Tootsille, joka teeskenteli olevansa sen
hyvä tuttu, ja syöksyi silmittömänä Towlinsonin kimppuun,
vilpittömän vannana siitä, että tuossa oli se vihollinen, jota hän oli
haukkunut koko ikänsä nurkan takaa koskaan häntä näkemättä, oli
Florence siihen niin tyytyväinen kuin se olisi ollut oikea viisauden
ihme.
Siitä, mitä hän lisäsi ja oli kuullut alakerrassa, kävi ilmi, että rouva
Chick oli ehdottanut majuria Dombeyn matkatoveriksi ja että
Dombey jonkin verran epäröityään oli kutsunut hänet.
Hänen säälivä äänensävynsä kosketti sitä kieltä, jota niin usein oli
kovakouraisesti kosketeltu, mutta jonka värähdystä ei kukaan ollut
kuullut. Yksin jäätyään Florence painoi päänsä kättään vasten ja
pannen toisen kätensä kiihkeästi kohoilevalle povelleen päästi
surunsa valloilleen.
Ilta oli kostea, ja alakuloinen sade pieksi yksitoikkoisesti kohisten
akkunoita. Hidas tuuli puhalsi ja kiersi valittaen taloa kuin tuntien
tuskaa tai surua. Kimeä ääni vaikerteli puissa. Hänen istuessaan siinä
itkemässä tuli myöhä ja kirkontornista kuului juhlallinen ja kaamea
keskiyön soitto.
Kun ovi oli auki, vaikka ainoastaan niinkin vähän, tuntui siinä
olevan toivoa. Hänestä oli rohkaisevaa nähdä sisältä valonsäde, joka
pujahti ulos raosta pimeään ja levisi kultaisena juovana
marmorilattialle. Hän kääntyi takaisin tuskin tietäen mitä teki, mutta
häntä veti sinne sydämessä asuva rakkaus ja koettelemus, joka oli
kohdannut heitä kumpaakin, mutta josta he eivät olleet toisilleen
mitään puhuneet. Ja niin hän astui sisään vapisevat kädet vähän
kohotettuina.
"Mikä nyt on hätänä?" kysyi hän yrmeästi. "Miksi sinä tulet tänne?
Mikä on pelästyttänyt sinua?"
Jos jokin peloitti tyttöä, niin juuri nuo kasvot, jotka olivat häneen
suunnattuina. Nuoren tytön rinnassa palava rakkaus hyytyi jääksi
tuollaisen katseen edessä, ja niin hän seisoi tuijottaen isäänsä kuin
kivettyneenä.
"En, Wally, en", vastasi ukko, "kaikki, mitä kuulen neiti Dombeystä,
kun hän on jäänyt niin yksin, kirjoitan sinulle. Mutta pelkäänpä, ettei
siitä tule paljon, Wally."
"Kuuleppas, eno, mitä nyt kerron sinulle", virkkoi Walter hetken
epäröityään, "olen juuri käynyt siellä".
"Ja lisäsin", jatkoi Walter, "että jos hän — Susan nimittäin — voisi
joskus ilmoittaa sinulle joko itse tai rouva Richardsin kautta tai kenen
muun avulla tahansa, joka menisi tästä ohitse, että neiti Dombey on
hyvässä voinnissa, tuntuisi se sinusta hyvin ystävälliseltä, ja sinä
kirjoittaisit siitä minulle, ja minäkin olisin siitä hyvin kiitollinen. No
niin, usko pois, eno", lisäsi Walter, "että tuskin ollenkaan nukuin
viime yönä ajatellessani sitä vierailua, enkä voinut päästä selville,
oliko minun tehtävä niin vai ei. Ja kuitenkin olen varma, että minua
kehoitti siihen vilpitön tunne ja että olisin ollut kovin onneton
jälkeenpäin, jollen olisi siellä käynyt."
"Niin että jos joskus tapaat hänet", virkkoi Walter, "tarkoitan nyt
neiti Dombeyta — ja kuka tietää, ehkä sinä tapaatkin! — niin kerro
hänelle, kuinka suuri minun myötätuntoni häntä kohtaan on, kuinka
paljon ajattelin häntä täällä ollessani, kuinka puhuin hänestä
kyyneleet silmissä, eno, tänä viimeisenä iltana ennen lähtöäni. Kerro
hänelle, että sanoin aina muistavani hänen ystävällisen käytöksensä,
hänen kauniit kasvonsa, hänen suloisen luonteensa, joka on paras
kaikesta. Ja koska en ottanut näitä naisen tai nuoren neidon jaloista,
vaan pienen viattoman lapsen", jatkoi Walter, "niin sano hänelle,
eno, että minä pidin nämä kengät — hän kyllä muistaa, kuinka usein
ne silloin putosivat — ja vein ne mukanani muistoksi."