Donald C. Mitchell, D. Lawrence Bunk,
MD Jr, MD
#{149} Simon #{149} Matthew
Vinitski, PhD D. Rifkin,
#{149} Stephen
Saponaro,
Liver and Pancreas: Improved Ti Contrast by Shorter Echo Fat Suppression at 1.5 T’ Ti-weighted spin-echo magnetic resonance (MR) images have had limited soft-tissue contrast at 1.5 T. The authors investigated the effects of echo-time (TE) minimization and fat suppression on MR images of the liver and pancreas. Two sets of MR images were obtained with identical repetition times and other parameters. In 10 subjects with seven liver lesions, images with TEs of 20 and
12 msec
were
compared.
In
18 additional subjects with seven liver lesions and five pancreatic cancinomas, images with identical TEs but with and without fat suppression were compared. Contrast-tonoise ratios (CNRs) were greater with a TE of 12 msec than with a TE of 20 msec for liver versus spleen (7.6 vs 4.9, P = .014) and liver versus lesion (6.9 vs 3.9, P .031). In patients without fatty liver, CNR for six lesions versus liver was greater (9.5 vs 6.0, P = .014) with fat suppression. CNR between glandular pancreas and cancer was most conspicuous with fat suppression, but fat planes were less distinct. Minimization of TE improves Tiweighted images significantly. Fat suppression also improves CNR, but the disadvantages of fat suppression do not allow elimination of conventional Ti-weighted images. Index
terms:
Magnetic ment
Liver,
resonance
(MR),
#{149} Magnetic
quences
resonance
#{149} Pancreas,
Radiology
I
From
MR
1991;
the
ferson 11th
Department
Medical and Sansom
Received
June
C
22. RSNA,
761.1214
contrast
enhance-
(MR), studies,
pulse
Se-
770.1214
of Radiology,
Resonance University College, Sts, 15,
7; revision
August
studies,
178:67-71
sion of Magnetic Thomas Jefferson
gust
MR
Address 1991
Divi-
Imaging, Hospital
The and
Jef-
10th Fl, Main Bldg. Philadelphia, PA 19107.
1990;
received reprint
revision August requests
requested 20;
Au-
accepted to D.G.M.
W
proper
ITH
gies
BA
attention
to strate-
of motion
artifact, high-quality images of the abdomen can be obtained at 1.5 T (1-3). At this field strength, spinecho magnetic resonance (MR) images with long repetition time/echo time (TRITE) appear to be at least as sensitive as midfield MR images or contrast material-enhanced computed tornognams (4,5). These images take a long time to acquire, however, and can be degraded by motion. Pe-
nipheral
hepatic may
that
of fat, and
lesions
have
mimicked
by
tic vessels
or adenop-
intensity
small
similar
lesions
high-signal
when is used
artifact.
In addition,
to
may
be
intrahepa-
gradient
nulhing
moment
to minimize
motion
depiction
of the
normal and abnormal pancreas remains suboptirnah on long TRITE images (6). For the reasons stated above, Tiweighted images of the abdomen are necessary at 1.5 T, even though Ti contrast between most tissues decreases at high field strength (7,8). Ti contrast can be increased with invension recovery (9,10), but the long irnaging time, low signal-to-noise ratio, vulnerability to motion artifact, and limited number of sections imaged
have been investigators ing results
problems. Although some have reported promisof techniques involving
spoiled gradient-echo images obtamed with very short TE and breath holding (1 1), these have a low signalto-noise ratio. Thus, most radiologists continue to obtain spin-echo images in spite of limited contrast and sensitivity for lesions. Ti contrast on spin-echo images is reduced by contamination of tissue contrast from T2 differences. This reduction in contrast can be minimized by reducing TE. The simplest method of reducing TE is to decrease gradient duration while increasing gradient
strength This
for a given method
Tasciyan,
PhD
Spin-Echo Time and
for suppression
athy
#{149} Talin
MD
requires
field good
of view. compen-
sation for eddy currents. The sampling bandwidth must also be increased, which reduces the signal-tonoise ratio. For these reasons, there a lower limit for optimizing TE beyond which further reduction may
reduce
the
contrast
signal-to-noise
ratio
is
and
(12).
Contrast
on Ti-weighted
images
is
also limited by the high signal of fat, which contributes to image noise and reduces the dynamic range of water signal. One potential solution to these problems is suppression of signal from fat by nadiofrequency saturation (13,14). To determine whether shorter TE and fat suppression could improve Ti-weighted images of the liven and pancreas at 1.5 T, we investigated the effects of decreasing TE and using fat suppression while
holding
all other
parameters
con-
stant.
PATIENTS
AND
Two nearly ed images utes
METHODS
identical
were
of each
sets of Ti-weight-
obtained
other
within
for
each
30 mm-
of 33 subjects.
There were three comparisons. of 12 and 20 msec were compared subjects. fat
Then,
suppression
images
with
were
compared
jects. Finally, two sets images, acquired with opposed-phase
with
l.5-T
imager
Medical for
each
other
with
3.2
Systems,
thickness sections
(28-36
cm2),
(four), msec) tion Two
and were
were
in five
TE
=
echo
was
used
location,
of signals
averaged
time (400 or 500 Receiver attenua-
separately, acquisitions
CNR time,
A GE
(Signa;
however. were re-
quired to cover the liver, resulting tat imaging time of 6.9 minutes
tio,
and
corn-
subjects.
software
section
number
Abbreviations:
18 sub-
(7 or 10 mm), gap be(2 mm), field of view
repetition identical.
was optimized interleaved
without in
Milwaukee)
all examinations. For alt comparisons,
section tween
and
of fat-suppressed the in-phase
techniques,
pared
First, TEs in 10
TR
per
contrast-to-noise repetition
in a tose-
ratime.
67
quence. Opposed-phase Ti-weighted imaging without fat suppression but with identical TR and TE within 2 msec was also performed in all patients. Opposedphase images were obtained by changing the gradient timing so that, when the echo is formed, fat and water magnetization are antiparallel (15). These images were used to identify fatty infiltration, based on comparison of the relative intensity of the liver and spleen on inphase and opposed-phase images (15,16). Motion artifact was reduced for all Tiweighted images by high-frequency respiratory-ordered phase encoding (Exorcist; GE Medical Systems) and spatial presaturation of bbood flowing into the imaging volume. Finally, T2-weighted images (TR/TE 2,500/100 rnsec; imaging time = 11.2 minutes) were obtained for all patients at identical locations with use of an identical field of view and section thickness and gap and the average of two signals. Motion artifact was suppressed with respiratory-ordered phase encoding, nubling the first moment of the gradient, decreasing the interecho interval (first echo 50 msec) (2), and spatial presaturation.
For 10 subjects, one set of Ti-weighted images was obtained with the manufacturer’s minimum TE of 20 msec (sampling bandwidth 32 kHz), and another set was obtained with a reduced TE of 12 msec (45 kHz), both with symmetric data sampling. On the imaging system used, the receiver bandwidth is fixed at 32 kHz. Thus, 13% of each side of the image along the frequency-encoding axis was attenuated (Fig 1). This problem can be corrected with appropriate modifications of hardware. On images of two persons, attenuation of the lateral aspects of the body was prevented by changing the axis of frequency encoding to anteroposterior for both sequences. This change resulted in increased degradation of the image of the liver by motion on the left side of the abdomen, however (Fig 2). In the second comparison, Ti-weighted imaging was performed with and without fat suppression in 18 subjects. Signal from fat was suppressed before each excitation pulse by a spatially nonsebective 90#{176} cornposite radio-frequency pulse tuned mdividualby to excite triglyceride within the imaging volume. This pulse was followed by a spoiling gradient pulse to dephase the resulting lipid transverse magnetization. The remaining water magnetization was then excited to create an image dominated by water. In five of the 18 subjects, fat suppression was augmented with opposed-phase imaging after saturation of fat. The rationale for this has been presented by Listerud et al (17). Olefinic fatty acids, which make up approximately 5% of adipose tissue, have a resonant frequency similar to that of water and are thus not affected by the saturation pulse. If tuning is not perfect, some residual triglyceride will remain as well. In an opposed-phase image, the phases of obefinic acids and trigbyceride will be opposite, resulting in signal
68
#{149} Radiology
cancellation and thus increasing the net suppression of lipid signal. Since the minimum TE of opposedphase images is greater than that of inphase images, the TE of fat-suppressed opposed-phase images (14 msec) was greater than that of the comparable nonfat-suppressed images (12 msec) in the five subjects from whom they were obtamed. For the 13 subjects whose fat-suppressed images were in phase, the TEs were identical for both sets of images (12 msec for 1 1 subjects, 22 msec for two subjects).
In
five
additional
persons,
fat-sup-
pressed in-phase (TE 12 msec) and opposed-phase (TE = 14 msec) images of the pancreas were compared visually for the degree of fat suppression. A contrast-to-noise ratio (CNR) for liver versus spleen was calculated as a model for contrast between liver and lesion (i8) by dividing the difference in signal between the two by the standard deviation of background noise anterior to the abdomen. Thus, motion-induced and system noise are accounted for in the resulting CNR. The three regions of interest (liver, spleen, background) were identical for both images within each comparison and were chosen on the same section, avoiding vessels and regions where signal was attenuated by the frequency filter. The baseline value 1,024, unique to this systern, was subtracted from each intensity measurement. Ten subjects had a total of 14 focal hepatic lesions. For these subjects, CNR was calculated for liver versus lesion by the method described above. This led to comparison of i2-msec versus 20-msec TEs for seven lesions (three metastases and four benign lesions) and fat suppression versus non-fat suppression for seven lesions (four metastases, two benign lesions, and one focal nodular hyperplasia). Metastases and focal nodular hyperplasia were histologically proved, whereas benign lesions (hemangiomas or cysts) were presumed based on appearance at MR imaging and lack of a known malignancy. Images were also analyzed subjectively for contrast and motion artifact. The contrast and
of the bowel
pancreas relative to the liver was subject to special atten-
tion. In five subjects with pancreatic mom, conspicuity of the tumor and marcation relative to the remainder pancreas were assessed. CNR between cancer and glandular pancreas and was measured. Statistical significance determined with the Wilcoxon sign test, a nonparametric test.
tuits deof the
Figure
was rank
Comparison
of a TE of 20 msec
Figure
2. Comparison of a TE of 20 msec (top) and a TE of 12 msec (bottom) in a patient with a metastasis from colon carcinoma. The frequency-encoding axis has been switched to anteropostenior, avoiding attenuation
fat
1.
(top) and a TE of 12 msec (bottom) in a patient with a metastasis from colon carcinoma. The lesion is much more obvious with a TE of 12 msec (arrows). Note attenuation by the frequency filter of the left and right portions of the image at bottom.
of
the
left
and
right
portions
of
the
image. Motion artifact from the left side of the abdomen degrades the image of the liven, however, and obscures the lesion with a TE of 20 msec. The lesion is obvious with a TE of 12 msec, however (arrows).
RESULTS CNR was greater with a TE of 12 msec than with a TE of 20 msec for liven versus spleen (7.6 ± 4.9 vs 4.9 ± 2.9, P = .014) and liver versus lesion (6.9 ± 5.0 vs 3.9 ± 3.1, P = .031) (Figs 1-3). Motion artifact was also less apparent, consistent with the shorter TE. No significant distortion of irn-
ages (which would have suggested artifacts related to eddy currents) was noted. Motion artifact in images with fat suppression was similar to on less than that in images without fat suppression for all 18 subjects from whom such images were obtained. Fatty liver was detected in five of
January
1991
15
#{149}
10
#{149}
15
#{149}
a
#{149} #{149}
I0 0
5
A
#{149}
A
8
9
5
A
#{149}
1
2
A
I
2
3
5
4
6
7
10
3
Lesion
.
5
6
7
Lesion
b. Figure
3.
Scatter
diagrams
comparing
contrast
for spleen
versus
liver
(a) and
lesion
versus
liven (b) with a TE of 20 msec (X) versus a TE of 12 msec (#{149}).Contrast was better with a TE of 12 msec in eight of 10 comparisons for spleen versus liven and in five of seven comparisons for lesion versus liven. Lesions 1-4 were presumed hemangiomas on cysts, whereas lesions 5-7 were metastases from cobonic carcinoma.
20
#{149}
15
0
Figure 4. Images of a patient with fatty liven (TE = 12 msec). A presumed hemangioma or cyst (arrow) is obvious in the image obtamed without fat suppression (top) but is obscured 0
#{149}
10
because #{149}
(bottom). This the substantial
0
0 0
S
S
2
3
4
5
section
Scatter
diagram
comparing
T2-weighted
Figure 7. Axial images of a patient colonectal metastasis. Ti weighting
178
#{149} Number
1
be
of the liver
image.
of
Fat suppression
at the edges of the is distant from the
volume,
causing
of triglyceride
different
in spite artifact
from
the
the
in this frequency
of
pulse.
con-
images
with
with
opposed-phase technique of fat suppression (top) is compared with T2 weighting (bottom). The lesion is distinct on the Tiweighted image (arrow) but subtle on the T2-weighted image, except for the center, which might represent central necrosis.
Volume
to
the fat saturation
trast for lesion versus liver with conventionat Ti-weighted images (X), fat-suppressed Ti-weighted images (#{149} ), and T2-weighted images (0). Lesions 1-4 are cobonectal metastases, lesion 5 is focal nodular hyperplasia, and lesions 6 and 7 are presumed hemangiomas or cysts. * = fatty liven, causing reduced contrast on the fat suppression image. Excluding the images of the patient with fatty liver, fat-suppressed Ti-weighted images had greater contrast than conventional Tiweighted images in all six comparisons and
greaten contrast than four of six comparisons.
frequency
fat suppression occurs in motion
especially this section
of the imaging
resonant
7.
6
Lesion 5.
with
intensity
obscuration reduction
is suboptimal, field, because
center
Figure
image
in the fat-suppressed
0 #{149}
1
in the
of decreased
not the Figure (top) with tastasis
in
a
the
6.
Comparison
and fat-suppressed a TE of 20 msec from
colon
of conventional (bottom) in a patient
carcinoma.
images with a meThe
lesion
is
obscured on the conventional image by motion-induced artifact but is obvious with fat suppression (arrow). Note the small high-intensity mass in the night kidney suggestive of a hemorrhagic cyst.
these 18 subjects by comparison of in-phase and opposed-phase non-fatsuppressed images; the latter sequence depicts fatty liver as less intense than spleen because of phase cancellation between fat and water (15,16). In these five subjects, CNR between liven and spleen was decreased by fat suppression and a liver lesion was obscured in spite of substantially reduced motion artifact (Fig 4). In the 13 subjects without fatty liver, CNR was significantly greaten with fat suppression for liven yensus lesion (9.5 ± 6.7 vs 6.0 ± 3.7; P .0i4) (Figs 5, 6) and slightly greater for liver versus spleen (6.9 vs 5.4; P .077). Fat suppression was occasionalhy suboptimal at the periphery of the liver, especially at the top and bottorn sections (Fig 4), and occasionally resulted in water suppression at the lowest sections. These limitations did
interfere with depiction of any of seven lesions in this series. Among patients without evidence of fatty liver, fat-suppressed Tiweighted images and T2-weighted images had similar CNRs (9.5 ± 6.7 vs 10.6 ± 4.1; P .66) (Figs 5, 7). CNRs for the four metastases were 6.2 ± 3.5 for conventional Ti-weighted images, 10.3 ± 7.3 for fat-suppressed Ti-weighted images, and 9.5 ± 4.8 for T2-weighted images. Fat suppression was effective for the penipancreatic region in all 18 subjects for whom such images were obtained. On these images, the pancreas had higher signal intensity than all other solid tissues and was at least as intense as bowel. Fat suppression in opposed-phase images was superior to that in in-phase images of the five individuals for whom both were obtained (Fig 8). Because the signal from the glandular pancreas is so much greaten than that from nesiduah fat after presatunation, even a markedly fatty pancreas was hypenintense on fat-suppressed opposedphase images (Fig 9). Pancreatic cancer was less intense than the glandulan pancreas and was more conspicuous with fat suppression (Fig iO). CNR was greaten for pancreas versus tumor with fat suppression (6.4 ± 2.6 vs 3.5 ± 1.2) but less for tumon versus fat (15.7 ± 19.8 vs 28.8 ±
Radiology
#{149} 69
15.0). These differences tisticalhy significant.
were
not
sta-
DISCUSSION Our data show for Ti-weighted proved substantially and We
that, images
by suppressing reduced TE by
at 1.5 T, CNR can be imby reducing TE
signal decreasing
from
fat. gradi-
ent duration and sampling time. As independent factors, reducing sampling time decreases signal-to-noise ratio, whereas reducing TE increases signal-to-noise ratio, minimizes contamination from T2 differences, and decreases motion artifact. The net effect of reducing TE in our series was a significant increase in CNR between liver and spleen and between liver and lesion. The only disadvantage of shorten TE in our series was attenuation at the edges of the image by the hardwane radio-frequency filter. Appropniate modifications of hardware can eliminate this problem. It is possible that further reduction of TE by asymmetric echo sampling, whether alone on in combination with decreased gradient duration, may yield even better results. Regardless of the method by which it is accomplished, TE should be minimized on most if not all Ti-weighted images of the abdomen at 1.5 T. Ti differences between most tissues can also be depicted better if signal from fat is suppressed. Because most signal in conventional Tiweighted images of the abdomen is from fat, much less signal is detected by the receiver when signal from fat is suppressed. This allows receiver gain to be increased, improves dynamic range, and allows better delineation of differences between other tissues. In addition, fat suppression reduces artifact caused by moving fat. Photography of the images also can be optimized to depict subtle differences in contrast between nonfatty tissues. Fat suppression was especially helpful for Ti-weighted images of the pancreas. Even at 1 .5 T, contrast between glandular pancreas and tumon appears to be better on Tiweighted than on T2-weighted images (6), probably because the desmoplastic reaction to most pancreatic carcinoma limits T2 relaxation times. When signal from fat was suppressed, the pancreas became brighten than all other solid tissues and pancreatic tumors were depicted as defects of bow signal intensity. We recommend using conventional Ti70
#{149} Radiology
a. Figure tional
b. 8.
Axial Ti-weighted images TE of 12 msec (a), fat-suppressed
TE of 14 msec (c) are compared. With sue in the image. Note the superiority in C.
a. Figure ventional
texture. tissues,
of the pancreatic TE of i2 msec
head (b),
c. in a healthy
and
Axial image
images
with
of
an
In the opposed-phase, in spite
fat suppression water within
of suppression
(c), the pancreas the same voxels.
c.
elderly
TE of 12 msec
Conven-
opposed-phase
fat suppression, the pancreas becomes the brightest tisof fat suppression with the opposed-phase technique
b. 9.
volunteer.
fat-suppressed
man
(a),
with
the
pancreas
fat-suppressed of its fatty
fatty
image components.
has decreased
weighted sequences as well to differentiate fat from other tissues that have low signal intensity on fat-suppressed images. Unlike reduction of TE, fat suppression has significant disadvantages and cannot replace conventional Ti-weighted images for most abdominal applications. Methods of fat suppression that depend on the chemical shift between fat and water require a homogeneous magnetic field. With present technology, it is difficult to achieve a homogeneous magnetic field throughout a large volume of interest. With our system, fat suppression was usually effective near the center of the volume of intenest, such as for the pancreas. Fat suppression was less effective at the periphery of the liver, however, especially at the top and bottom sections. Even worse, water suppression of the liven on bottom sections occunred occasionally, which potentialhy could obscure lesions. In the future, however, improved magnetic field homogeneity may increase the reliability of fat suppression for hepatic imaging. Contrast between fatty liven and lesion can be reduced by fat suppres-
intensity
infiltration
has
high
of
signal
(b), the pancreas In the
caused
the
pancreas.
intensity
In
and
is brighter
opposed-phase
by the presence
the
con-
a marbled
than image
other without
of fat and
Figure 10. Axial images (TE 12 msec) obtamed without (top) and with (bottom) fat suppression of a patient with carcinoma of the pancreatic tail. Contrast between tumor (T) and glandular pancreas (P) is better with fat suppression (signal difference = 19 vs 6, signal intensity difference 1.2 vs ii, CNR = 3.8 vs 1.9), but the fat planes are less distinct.
sion. This phenomenon has been noted with opposed-phase Tiweighted images (i9). Therefore, we recommend obtaining conventional Ti-weighted images to survey the liver, reserving fat suppression for a
January
1991
problem-solving sequence to augment contrast or determine whether fat is present. Opposed-phase images could be obtained and compared with in-phase images to detect fatty liven, thereby aiding in determining whether an additional fat suppression sequence might be helpful. Among patients without fatty liven, CNR between liven and lesion was similar to that obtained with techniques that have been necognized as sensitive. In fact, there was no significant difference in liven-hesion CNR between fat-suppressed Tiweighted images (9.5 ± 6.7) and T2weighted images (10.6 ± 4.1). CNR between liven and metastases for these two techniques (10.3 and 9.5, respectively) was similar to CNR calculated by the same method and necently reported for images obtained at 0.6 T with use of TRITE of 275/14 msec and 10 signal averages (10.6 ± 6.1) (20). In conclusion, reducing TE can improve CNR significantly in abdorninah Ti-weighted images. Decreasing gradient duration and sampling time is one effective method of accomphishing this. Fat suppression can improve Ti contrast further, but we necommend obtaining conventional Tiweighted sequences as well. Fatsuppressed Ti-weighted imaging appears promising for studying the pancreas. Further studies comparing the efficacy of MR imaging and other
Volume
178
#{149} Number
1
modalities may be
for imaging indicated. U
the
pancreas
1 1.
Edelman Dynamic Cd-DTPA:
12.
Hendrick RE. Sampling time effects on signal-to-noise and contrast-to-noise ratios in spin-echo MRI. Magn Reson Imaging 1987; 5:31-37. Keller PJ, Hunter WW Jr. Schmalbrock P. Multisection fat-water imaging with chemical shift selective presaturation. Radiology 1987; 164:539-541. Szumowski J, Eisen JK, Vinitski 5, Haake PW, Plewes DB. Hybrid methods of chemical-shift imaging. Magn Reson Med
1989;
References 1.
2.
3.
Mitchell DC, Vinitski S. Principles of protocol optimization for MRI of the abdomen and pelvis. Crit Rev Diagn Imaging (in press). Mitchell DC, Vinitski 5, Bunk DL, Levy DW, Rifkin MD. Motion artifact reduction in MR imaging of the abdomen: gradient moment nutting versus respiratory-sorted phase encoding. Radiology 1988; 169:155160. Felmlee JP, Ehman RL. Spatial presaturation: a method for suppressing flow artifacts and improving depiction of vascular anatomy in MR imaging. Radiology 1987;
13.
14.
1989;
15. 16.
164:559-564.
4.
Rummeny E, Wernecke K, Bongartz C, et al. ROC analysis of high-field MR imaging versus CT for the detection of focal hepatic
5.
6.
7.
8.
lesions.
Fullerton
CD,
quency
dependence
Cameron
KL,
Ord
VA.
of magnetic
MR
imaging.
149:831-837. Reinig JW. Dwyer tases:
detection
with
AE. MR
1987;
DL,
Liver
imaging
18.
Fre-
reso-
AJR
AJ, Miller
CW, Chang
17.
Imaging
nance spin-lattice relaxation of protons in biological materials. Radiology 1984; 151:135-138. Mitchell DC, Burk DL, Vinitski 5, et al. The biophysical basis of tissue contrast in
JA, Adams
10.
Reson
1990; 8(suppl 1):57. Vassiliades VC, Foley WD, Alarcon J, et al. Hepatic metastases: CT versus MR imaging at 1.5 T. Radiology 1989; t73(P):175. Romano W, Vellet D, Bach D. Comparative evaluation of CT and MR imaging of pancreatic carcinoma. Radiology 1989; 173(P):3l7-318.
extracraniat
9.
Magn
Frank
19.
20.
RR, Siegel JB, Singer A, et al. MR imaging of the liver with initial clinical results. AJR
153:1213-1219.
9:379-388.
Dixon WT. Simple proton spectroscopic imaging. Radiology 1984; 153:189-194. Schertz LD, Lee JKT, Heiken JP, Molina PL, Totty WC. Proton spectroscopic imaging (Dixon method) of the liver: clinical utility. Radiology 1989; 173:401-405. Listerud J, Isaac C, Chan T, Lenkinski RE. Aliphatic/olefinic fat cancellation: optimization and extension (abstr). In: Book of abstracts: Society of Magnetic Resonance in Medicine 1990. Berkeley, Calif: Society of Magnetic Resonance in Medicine, 1990; 590. Stark DD, Wittenberg J, Edelman RR, et al. Detection of hepatic metastases: analysis of pulse sequence performance in MR imaging. Radiology 1986; 159:365-370. Stark DD, Wittenberg J, Middleton MS. Ferrucci JT Jr. Liver metastases: detection by phase contrast MR imaging. Radiology 1986; 158:327-332. Dousset M, Weissleder R, Hendrick RE, et al. Short TI inversion-recovery imaging of the liver: pulse-sequence optimization and comparison with spin-echo imaging. Radiology 1989; 171:327-333.
metasat 0.5
and 1.5 T. Radiology 1989; 170:149-153. Steinberg HV, Alarcon JJ, Bernardino ME. Focal hepatic lesions: comparative MR imaging at 0.5 and 1.5 T. Radiology 1990; 174: 153-156.
Radiology
#{149} 71
MD Jr, MD
#{149} Simon #{149} Matthew
Vinitski, PhD D. Rifkin,
#{149} Stephen
Saponaro,
Liver and Pancreas: Improved Ti Contrast by Shorter Echo Fat Suppression at 1.5 T’ Ti-weighted spin-echo magnetic resonance (MR) images have had limited soft-tissue contrast at 1.5 T. The authors investigated the effects of echo-time (TE) minimization and fat suppression on MR images of the liver and pancreas. Two sets of MR images were obtained with identical repetition times and other parameters. In 10 subjects with seven liver lesions, images with TEs of 20 and
12 msec
were
compared.
In
18 additional subjects with seven liver lesions and five pancreatic cancinomas, images with identical TEs but with and without fat suppression were compared. Contrast-tonoise ratios (CNRs) were greater with a TE of 12 msec than with a TE of 20 msec for liver versus spleen (7.6 vs 4.9, P = .014) and liver versus lesion (6.9 vs 3.9, P .031). In patients without fatty liver, CNR for six lesions versus liver was greater (9.5 vs 6.0, P = .014) with fat suppression. CNR between glandular pancreas and cancer was most conspicuous with fat suppression, but fat planes were less distinct. Minimization of TE improves Tiweighted images significantly. Fat suppression also improves CNR, but the disadvantages of fat suppression do not allow elimination of conventional Ti-weighted images. Index
terms:
Magnetic ment
Liver,
resonance
(MR),
#{149} Magnetic
quences
resonance
#{149} Pancreas,
Radiology
I
From
MR
1991;
the
ferson 11th
Department
Medical and Sansom
Received
June
C
22. RSNA,
761.1214
contrast
enhance-
(MR), studies,
pulse
Se-
770.1214
of Radiology,
Resonance University College, Sts, 15,
7; revision
August
studies,
178:67-71
sion of Magnetic Thomas Jefferson
gust
MR
Address 1991
Divi-
Imaging, Hospital
The and
Jef-
10th Fl, Main Bldg. Philadelphia, PA 19107.
1990;
received reprint
revision August requests
requested 20;
Au-
accepted to D.G.M.
W
proper
ITH
gies
BA
attention
to strate-
of motion
artifact, high-quality images of the abdomen can be obtained at 1.5 T (1-3). At this field strength, spinecho magnetic resonance (MR) images with long repetition time/echo time (TRITE) appear to be at least as sensitive as midfield MR images or contrast material-enhanced computed tornognams (4,5). These images take a long time to acquire, however, and can be degraded by motion. Pe-
nipheral
hepatic may
that
of fat, and
lesions
have
mimicked
by
tic vessels
or adenop-
intensity
small
similar
lesions
high-signal
when is used
artifact.
In addition,
to
may
be
intrahepa-
gradient
nulhing
moment
to minimize
motion
depiction
of the
normal and abnormal pancreas remains suboptirnah on long TRITE images (6). For the reasons stated above, Tiweighted images of the abdomen are necessary at 1.5 T, even though Ti contrast between most tissues decreases at high field strength (7,8). Ti contrast can be increased with invension recovery (9,10), but the long irnaging time, low signal-to-noise ratio, vulnerability to motion artifact, and limited number of sections imaged
have been investigators ing results
problems. Although some have reported promisof techniques involving
spoiled gradient-echo images obtamed with very short TE and breath holding (1 1), these have a low signalto-noise ratio. Thus, most radiologists continue to obtain spin-echo images in spite of limited contrast and sensitivity for lesions. Ti contrast on spin-echo images is reduced by contamination of tissue contrast from T2 differences. This reduction in contrast can be minimized by reducing TE. The simplest method of reducing TE is to decrease gradient duration while increasing gradient
strength This
for a given method
Tasciyan,
PhD
Spin-Echo Time and
for suppression
athy
#{149} Talin
MD
requires
field good
of view. compen-
sation for eddy currents. The sampling bandwidth must also be increased, which reduces the signal-tonoise ratio. For these reasons, there a lower limit for optimizing TE beyond which further reduction may
reduce
the
contrast
signal-to-noise
ratio
is
and
(12).
Contrast
on Ti-weighted
images
is
also limited by the high signal of fat, which contributes to image noise and reduces the dynamic range of water signal. One potential solution to these problems is suppression of signal from fat by nadiofrequency saturation (13,14). To determine whether shorter TE and fat suppression could improve Ti-weighted images of the liven and pancreas at 1.5 T, we investigated the effects of decreasing TE and using fat suppression while
holding
all other
parameters
con-
stant.
PATIENTS
AND
Two nearly ed images utes
METHODS
identical
were
of each
sets of Ti-weight-
obtained
other
within
for
each
30 mm-
of 33 subjects.
There were three comparisons. of 12 and 20 msec were compared subjects. fat
Then,
suppression
images
with
were
compared
jects. Finally, two sets images, acquired with opposed-phase
with
l.5-T
imager
Medical for
each
other
with
3.2
Systems,
thickness sections
(28-36
cm2),
(four), msec) tion Two
and were
were
in five
TE
=
echo
was
used
location,
of signals
averaged
time (400 or 500 Receiver attenua-
separately, acquisitions
CNR time,
A GE
(Signa;
however. were re-
quired to cover the liver, resulting tat imaging time of 6.9 minutes
tio,
and
corn-
subjects.
software
section
number
Abbreviations:
18 sub-
(7 or 10 mm), gap be(2 mm), field of view
repetition identical.
was optimized interleaved
without in
Milwaukee)
all examinations. For alt comparisons,
section tween
and
of fat-suppressed the in-phase
techniques,
pared
First, TEs in 10
TR
per
contrast-to-noise repetition
in a tose-
ratime.
67
quence. Opposed-phase Ti-weighted imaging without fat suppression but with identical TR and TE within 2 msec was also performed in all patients. Opposedphase images were obtained by changing the gradient timing so that, when the echo is formed, fat and water magnetization are antiparallel (15). These images were used to identify fatty infiltration, based on comparison of the relative intensity of the liver and spleen on inphase and opposed-phase images (15,16). Motion artifact was reduced for all Tiweighted images by high-frequency respiratory-ordered phase encoding (Exorcist; GE Medical Systems) and spatial presaturation of bbood flowing into the imaging volume. Finally, T2-weighted images (TR/TE 2,500/100 rnsec; imaging time = 11.2 minutes) were obtained for all patients at identical locations with use of an identical field of view and section thickness and gap and the average of two signals. Motion artifact was suppressed with respiratory-ordered phase encoding, nubling the first moment of the gradient, decreasing the interecho interval (first echo 50 msec) (2), and spatial presaturation.
For 10 subjects, one set of Ti-weighted images was obtained with the manufacturer’s minimum TE of 20 msec (sampling bandwidth 32 kHz), and another set was obtained with a reduced TE of 12 msec (45 kHz), both with symmetric data sampling. On the imaging system used, the receiver bandwidth is fixed at 32 kHz. Thus, 13% of each side of the image along the frequency-encoding axis was attenuated (Fig 1). This problem can be corrected with appropriate modifications of hardware. On images of two persons, attenuation of the lateral aspects of the body was prevented by changing the axis of frequency encoding to anteroposterior for both sequences. This change resulted in increased degradation of the image of the liver by motion on the left side of the abdomen, however (Fig 2). In the second comparison, Ti-weighted imaging was performed with and without fat suppression in 18 subjects. Signal from fat was suppressed before each excitation pulse by a spatially nonsebective 90#{176} cornposite radio-frequency pulse tuned mdividualby to excite triglyceride within the imaging volume. This pulse was followed by a spoiling gradient pulse to dephase the resulting lipid transverse magnetization. The remaining water magnetization was then excited to create an image dominated by water. In five of the 18 subjects, fat suppression was augmented with opposed-phase imaging after saturation of fat. The rationale for this has been presented by Listerud et al (17). Olefinic fatty acids, which make up approximately 5% of adipose tissue, have a resonant frequency similar to that of water and are thus not affected by the saturation pulse. If tuning is not perfect, some residual triglyceride will remain as well. In an opposed-phase image, the phases of obefinic acids and trigbyceride will be opposite, resulting in signal
68
#{149} Radiology
cancellation and thus increasing the net suppression of lipid signal. Since the minimum TE of opposedphase images is greater than that of inphase images, the TE of fat-suppressed opposed-phase images (14 msec) was greater than that of the comparable nonfat-suppressed images (12 msec) in the five subjects from whom they were obtamed. For the 13 subjects whose fat-suppressed images were in phase, the TEs were identical for both sets of images (12 msec for 1 1 subjects, 22 msec for two subjects).
In
five
additional
persons,
fat-sup-
pressed in-phase (TE 12 msec) and opposed-phase (TE = 14 msec) images of the pancreas were compared visually for the degree of fat suppression. A contrast-to-noise ratio (CNR) for liver versus spleen was calculated as a model for contrast between liver and lesion (i8) by dividing the difference in signal between the two by the standard deviation of background noise anterior to the abdomen. Thus, motion-induced and system noise are accounted for in the resulting CNR. The three regions of interest (liver, spleen, background) were identical for both images within each comparison and were chosen on the same section, avoiding vessels and regions where signal was attenuated by the frequency filter. The baseline value 1,024, unique to this systern, was subtracted from each intensity measurement. Ten subjects had a total of 14 focal hepatic lesions. For these subjects, CNR was calculated for liver versus lesion by the method described above. This led to comparison of i2-msec versus 20-msec TEs for seven lesions (three metastases and four benign lesions) and fat suppression versus non-fat suppression for seven lesions (four metastases, two benign lesions, and one focal nodular hyperplasia). Metastases and focal nodular hyperplasia were histologically proved, whereas benign lesions (hemangiomas or cysts) were presumed based on appearance at MR imaging and lack of a known malignancy. Images were also analyzed subjectively for contrast and motion artifact. The contrast and
of the bowel
pancreas relative to the liver was subject to special atten-
tion. In five subjects with pancreatic mom, conspicuity of the tumor and marcation relative to the remainder pancreas were assessed. CNR between cancer and glandular pancreas and was measured. Statistical significance determined with the Wilcoxon sign test, a nonparametric test.
tuits deof the
Figure
was rank
Comparison
of a TE of 20 msec
Figure
2. Comparison of a TE of 20 msec (top) and a TE of 12 msec (bottom) in a patient with a metastasis from colon carcinoma. The frequency-encoding axis has been switched to anteropostenior, avoiding attenuation
fat
1.
(top) and a TE of 12 msec (bottom) in a patient with a metastasis from colon carcinoma. The lesion is much more obvious with a TE of 12 msec (arrows). Note attenuation by the frequency filter of the left and right portions of the image at bottom.
of
the
left
and
right
portions
of
the
image. Motion artifact from the left side of the abdomen degrades the image of the liven, however, and obscures the lesion with a TE of 20 msec. The lesion is obvious with a TE of 12 msec, however (arrows).
RESULTS CNR was greater with a TE of 12 msec than with a TE of 20 msec for liven versus spleen (7.6 ± 4.9 vs 4.9 ± 2.9, P = .014) and liver versus lesion (6.9 ± 5.0 vs 3.9 ± 3.1, P = .031) (Figs 1-3). Motion artifact was also less apparent, consistent with the shorter TE. No significant distortion of irn-
ages (which would have suggested artifacts related to eddy currents) was noted. Motion artifact in images with fat suppression was similar to on less than that in images without fat suppression for all 18 subjects from whom such images were obtained. Fatty liver was detected in five of
January
1991
15
#{149}
10
#{149}
15
#{149}
a
#{149} #{149}
I0 0
5
A
#{149}
A
8
9
5
A
#{149}
1
2
A
I
2
3
5
4
6
7
10
3
Lesion
.
5
6
7
Lesion
b. Figure
3.
Scatter
diagrams
comparing
contrast
for spleen
versus
liver
(a) and
lesion
versus
liven (b) with a TE of 20 msec (X) versus a TE of 12 msec (#{149}).Contrast was better with a TE of 12 msec in eight of 10 comparisons for spleen versus liven and in five of seven comparisons for lesion versus liven. Lesions 1-4 were presumed hemangiomas on cysts, whereas lesions 5-7 were metastases from cobonic carcinoma.
20
#{149}
15
0
Figure 4. Images of a patient with fatty liven (TE = 12 msec). A presumed hemangioma or cyst (arrow) is obvious in the image obtamed without fat suppression (top) but is obscured 0
#{149}
10
because #{149}
(bottom). This the substantial
0
0 0
S
S
2
3
4
5
section
Scatter
diagram
comparing
T2-weighted
Figure 7. Axial images of a patient colonectal metastasis. Ti weighting
178
#{149} Number
1
be
of the liver
image.
of
Fat suppression
at the edges of the is distant from the
volume,
causing
of triglyceride
different
in spite artifact
from
the
the
in this frequency
of
pulse.
con-
images
with
with
opposed-phase technique of fat suppression (top) is compared with T2 weighting (bottom). The lesion is distinct on the Tiweighted image (arrow) but subtle on the T2-weighted image, except for the center, which might represent central necrosis.
Volume
to
the fat saturation
trast for lesion versus liver with conventionat Ti-weighted images (X), fat-suppressed Ti-weighted images (#{149} ), and T2-weighted images (0). Lesions 1-4 are cobonectal metastases, lesion 5 is focal nodular hyperplasia, and lesions 6 and 7 are presumed hemangiomas or cysts. * = fatty liven, causing reduced contrast on the fat suppression image. Excluding the images of the patient with fatty liver, fat-suppressed Ti-weighted images had greater contrast than conventional Tiweighted images in all six comparisons and
greaten contrast than four of six comparisons.
frequency
fat suppression occurs in motion
especially this section
of the imaging
resonant
7.
6
Lesion 5.
with
intensity
obscuration reduction
is suboptimal, field, because
center
Figure
image
in the fat-suppressed
0 #{149}
1
in the
of decreased
not the Figure (top) with tastasis
in
a
the
6.
Comparison
and fat-suppressed a TE of 20 msec from
colon
of conventional (bottom) in a patient
carcinoma.
images with a meThe
lesion
is
obscured on the conventional image by motion-induced artifact but is obvious with fat suppression (arrow). Note the small high-intensity mass in the night kidney suggestive of a hemorrhagic cyst.
these 18 subjects by comparison of in-phase and opposed-phase non-fatsuppressed images; the latter sequence depicts fatty liver as less intense than spleen because of phase cancellation between fat and water (15,16). In these five subjects, CNR between liven and spleen was decreased by fat suppression and a liver lesion was obscured in spite of substantially reduced motion artifact (Fig 4). In the 13 subjects without fatty liver, CNR was significantly greaten with fat suppression for liven yensus lesion (9.5 ± 6.7 vs 6.0 ± 3.7; P .0i4) (Figs 5, 6) and slightly greater for liver versus spleen (6.9 vs 5.4; P .077). Fat suppression was occasionalhy suboptimal at the periphery of the liver, especially at the top and bottorn sections (Fig 4), and occasionally resulted in water suppression at the lowest sections. These limitations did
interfere with depiction of any of seven lesions in this series. Among patients without evidence of fatty liver, fat-suppressed Tiweighted images and T2-weighted images had similar CNRs (9.5 ± 6.7 vs 10.6 ± 4.1; P .66) (Figs 5, 7). CNRs for the four metastases were 6.2 ± 3.5 for conventional Ti-weighted images, 10.3 ± 7.3 for fat-suppressed Ti-weighted images, and 9.5 ± 4.8 for T2-weighted images. Fat suppression was effective for the penipancreatic region in all 18 subjects for whom such images were obtained. On these images, the pancreas had higher signal intensity than all other solid tissues and was at least as intense as bowel. Fat suppression in opposed-phase images was superior to that in in-phase images of the five individuals for whom both were obtained (Fig 8). Because the signal from the glandular pancreas is so much greaten than that from nesiduah fat after presatunation, even a markedly fatty pancreas was hypenintense on fat-suppressed opposedphase images (Fig 9). Pancreatic cancer was less intense than the glandulan pancreas and was more conspicuous with fat suppression (Fig iO). CNR was greaten for pancreas versus tumor with fat suppression (6.4 ± 2.6 vs 3.5 ± 1.2) but less for tumon versus fat (15.7 ± 19.8 vs 28.8 ±
Radiology
#{149} 69
15.0). These differences tisticalhy significant.
were
not
sta-
DISCUSSION Our data show for Ti-weighted proved substantially and We
that, images
by suppressing reduced TE by
at 1.5 T, CNR can be imby reducing TE
signal decreasing
from
fat. gradi-
ent duration and sampling time. As independent factors, reducing sampling time decreases signal-to-noise ratio, whereas reducing TE increases signal-to-noise ratio, minimizes contamination from T2 differences, and decreases motion artifact. The net effect of reducing TE in our series was a significant increase in CNR between liver and spleen and between liver and lesion. The only disadvantage of shorten TE in our series was attenuation at the edges of the image by the hardwane radio-frequency filter. Appropniate modifications of hardware can eliminate this problem. It is possible that further reduction of TE by asymmetric echo sampling, whether alone on in combination with decreased gradient duration, may yield even better results. Regardless of the method by which it is accomplished, TE should be minimized on most if not all Ti-weighted images of the abdomen at 1.5 T. Ti differences between most tissues can also be depicted better if signal from fat is suppressed. Because most signal in conventional Tiweighted images of the abdomen is from fat, much less signal is detected by the receiver when signal from fat is suppressed. This allows receiver gain to be increased, improves dynamic range, and allows better delineation of differences between other tissues. In addition, fat suppression reduces artifact caused by moving fat. Photography of the images also can be optimized to depict subtle differences in contrast between nonfatty tissues. Fat suppression was especially helpful for Ti-weighted images of the pancreas. Even at 1 .5 T, contrast between glandular pancreas and tumon appears to be better on Tiweighted than on T2-weighted images (6), probably because the desmoplastic reaction to most pancreatic carcinoma limits T2 relaxation times. When signal from fat was suppressed, the pancreas became brighten than all other solid tissues and pancreatic tumors were depicted as defects of bow signal intensity. We recommend using conventional Ti70
#{149} Radiology
a. Figure tional
b. 8.
Axial Ti-weighted images TE of 12 msec (a), fat-suppressed
TE of 14 msec (c) are compared. With sue in the image. Note the superiority in C.
a. Figure ventional
texture. tissues,
of the pancreatic TE of i2 msec
head (b),
c. in a healthy
and
Axial image
images
with
of
an
In the opposed-phase, in spite
fat suppression water within
of suppression
(c), the pancreas the same voxels.
c.
elderly
TE of 12 msec
Conven-
opposed-phase
fat suppression, the pancreas becomes the brightest tisof fat suppression with the opposed-phase technique
b. 9.
volunteer.
fat-suppressed
man
(a),
with
the
pancreas
fat-suppressed of its fatty
fatty
image components.
has decreased
weighted sequences as well to differentiate fat from other tissues that have low signal intensity on fat-suppressed images. Unlike reduction of TE, fat suppression has significant disadvantages and cannot replace conventional Ti-weighted images for most abdominal applications. Methods of fat suppression that depend on the chemical shift between fat and water require a homogeneous magnetic field. With present technology, it is difficult to achieve a homogeneous magnetic field throughout a large volume of interest. With our system, fat suppression was usually effective near the center of the volume of intenest, such as for the pancreas. Fat suppression was less effective at the periphery of the liver, however, especially at the top and bottom sections. Even worse, water suppression of the liven on bottom sections occunred occasionally, which potentialhy could obscure lesions. In the future, however, improved magnetic field homogeneity may increase the reliability of fat suppression for hepatic imaging. Contrast between fatty liven and lesion can be reduced by fat suppres-
intensity
infiltration
has
high
of
signal
(b), the pancreas In the
caused
the
pancreas.
intensity
In
and
is brighter
opposed-phase
by the presence
the
con-
a marbled
than image
other without
of fat and
Figure 10. Axial images (TE 12 msec) obtamed without (top) and with (bottom) fat suppression of a patient with carcinoma of the pancreatic tail. Contrast between tumor (T) and glandular pancreas (P) is better with fat suppression (signal difference = 19 vs 6, signal intensity difference 1.2 vs ii, CNR = 3.8 vs 1.9), but the fat planes are less distinct.
sion. This phenomenon has been noted with opposed-phase Tiweighted images (i9). Therefore, we recommend obtaining conventional Ti-weighted images to survey the liver, reserving fat suppression for a
January
1991
problem-solving sequence to augment contrast or determine whether fat is present. Opposed-phase images could be obtained and compared with in-phase images to detect fatty liven, thereby aiding in determining whether an additional fat suppression sequence might be helpful. Among patients without fatty liven, CNR between liven and lesion was similar to that obtained with techniques that have been necognized as sensitive. In fact, there was no significant difference in liven-hesion CNR between fat-suppressed Tiweighted images (9.5 ± 6.7) and T2weighted images (10.6 ± 4.1). CNR between liven and metastases for these two techniques (10.3 and 9.5, respectively) was similar to CNR calculated by the same method and necently reported for images obtained at 0.6 T with use of TRITE of 275/14 msec and 10 signal averages (10.6 ± 6.1) (20). In conclusion, reducing TE can improve CNR significantly in abdorninah Ti-weighted images. Decreasing gradient duration and sampling time is one effective method of accomphishing this. Fat suppression can improve Ti contrast further, but we necommend obtaining conventional Tiweighted sequences as well. Fatsuppressed Ti-weighted imaging appears promising for studying the pancreas. Further studies comparing the efficacy of MR imaging and other
Volume
178
#{149} Number
1
modalities may be
for imaging indicated. U
the
pancreas
1 1.
Edelman Dynamic Cd-DTPA:
12.
Hendrick RE. Sampling time effects on signal-to-noise and contrast-to-noise ratios in spin-echo MRI. Magn Reson Imaging 1987; 5:31-37. Keller PJ, Hunter WW Jr. Schmalbrock P. Multisection fat-water imaging with chemical shift selective presaturation. Radiology 1987; 164:539-541. Szumowski J, Eisen JK, Vinitski 5, Haake PW, Plewes DB. Hybrid methods of chemical-shift imaging. Magn Reson Med
1989;
References 1.
2.
3.
Mitchell DC, Vinitski S. Principles of protocol optimization for MRI of the abdomen and pelvis. Crit Rev Diagn Imaging (in press). Mitchell DC, Vinitski 5, Bunk DL, Levy DW, Rifkin MD. Motion artifact reduction in MR imaging of the abdomen: gradient moment nutting versus respiratory-sorted phase encoding. Radiology 1988; 169:155160. Felmlee JP, Ehman RL. Spatial presaturation: a method for suppressing flow artifacts and improving depiction of vascular anatomy in MR imaging. Radiology 1987;
13.
14.
1989;
15. 16.
164:559-564.
4.
Rummeny E, Wernecke K, Bongartz C, et al. ROC analysis of high-field MR imaging versus CT for the detection of focal hepatic
5.
6.
7.
8.
lesions.
Fullerton
CD,
quency
dependence
Cameron
KL,
Ord
VA.
of magnetic
MR
imaging.
149:831-837. Reinig JW. Dwyer tases:
detection
with
AE. MR
1987;
DL,
Liver
imaging
18.
Fre-
reso-
AJR
AJ, Miller
CW, Chang
17.
Imaging
nance spin-lattice relaxation of protons in biological materials. Radiology 1984; 151:135-138. Mitchell DC, Burk DL, Vinitski 5, et al. The biophysical basis of tissue contrast in
JA, Adams
10.
Reson
1990; 8(suppl 1):57. Vassiliades VC, Foley WD, Alarcon J, et al. Hepatic metastases: CT versus MR imaging at 1.5 T. Radiology 1989; t73(P):175. Romano W, Vellet D, Bach D. Comparative evaluation of CT and MR imaging of pancreatic carcinoma. Radiology 1989; 173(P):3l7-318.
extracraniat
9.
Magn
Frank
19.
20.
RR, Siegel JB, Singer A, et al. MR imaging of the liver with initial clinical results. AJR
153:1213-1219.
9:379-388.
Dixon WT. Simple proton spectroscopic imaging. Radiology 1984; 153:189-194. Schertz LD, Lee JKT, Heiken JP, Molina PL, Totty WC. Proton spectroscopic imaging (Dixon method) of the liver: clinical utility. Radiology 1989; 173:401-405. Listerud J, Isaac C, Chan T, Lenkinski RE. Aliphatic/olefinic fat cancellation: optimization and extension (abstr). In: Book of abstracts: Society of Magnetic Resonance in Medicine 1990. Berkeley, Calif: Society of Magnetic Resonance in Medicine, 1990; 590. Stark DD, Wittenberg J, Edelman RR, et al. Detection of hepatic metastases: analysis of pulse sequence performance in MR imaging. Radiology 1986; 159:365-370. Stark DD, Wittenberg J, Middleton MS. Ferrucci JT Jr. Liver metastases: detection by phase contrast MR imaging. Radiology 1986; 158:327-332. Dousset M, Weissleder R, Hendrick RE, et al. Short TI inversion-recovery imaging of the liver: pulse-sequence optimization and comparison with spin-echo imaging. Radiology 1989; 171:327-333.
metasat 0.5
and 1.5 T. Radiology 1989; 170:149-153. Steinberg HV, Alarcon JJ, Bernardino ME. Focal hepatic lesions: comparative MR imaging at 0.5 and 1.5 T. Radiology 1990; 174: 153-156.
Radiology
#{149} 71
