763

Detection of Hepatic Metastases:

Comparison of Contrast-Enhanced CT,
Unenhanced MR Imaging, and Iron Oxide-
Enhanced MR Imaging

Christian J. Fretz1’2 Diagnostic accuracy of contrast-enhanced CT, unenhanced MR imaging, and MR

David D. Stark1 images enhanced with superparamagnetic iron oxide was evaluated in 10 patients with
Charles E. Metz3 histologically proved hepatic metastases. First, diagnostic performance of the imaging
technique with respect to the ability of radiologists to recognize the presence or absence
Guillermo
of a metastasis was measured by using receiver-operating-characteristic (ROC) analysis
Ralph Weissleder1
of single images. Second, the total number of lesions (N = 108) detected by “complete”
Jong-Her Shen3 cT and MR examinations was counted. Finally, lesion-liver contrast-to-noise ratios (CNR)
Jack Wittenberg1 were measured in all MR sequences. The area under the ROC curve was .67 ± .03 for
Joseph Simeone1 contrast-enhanced CT, .81 ± .07 for the unenhanced SE 260/14 sequence, and .92 ±
Joseph T. Ferrucci1 .01 for the iron oxide-enhanced SE 1500/40 sequence. The enhanced SE 1500/40
sequence yielded significantly (p < .005) greater accuracy than did contrast-enhanced
cT. The same sequence detected significantly (p < .05) more lesions than all other
imaging techniques (19% more than the best unenhanced MR sequence and 36% more
than contrast-enhanced ci). The enhanced SE 1500/40 sequence also yielded the
highest CNR value (19.5 ± 10.2) of all MR sequences.
These results indicate that iron oxide-enhanced MR imaging is a superior imaging
technique for the detection of hepatic lesions.

AJR 155:763-770, October 1990

Detection of hepatic metastases is critical to treatment planning at the time of

diagnosis and during follow-up of a large number of cancer patients [1 -3]. Fur-
thermore, hepatic resection for cancer has shown increased survival rates for
selected patients [4-9]. Therefore, not only the presence or absence of hepatic
lesions (in an all-or-none fashion), but also their number, location, and size may
influence therapy and patients’ outcome [3, 4, 7, 10-12].
CT has been considered the gold standard for detection of focal hepatic lesions
because it has greater sensitivity and specificity than sonography
or scintigraphy
Received February 26, 1 990; accepted after re- [1 3]. Various contrast enhancement techniques [1 4-1 8] have been proposed to
vision May 23, 1990. improve furtherthe diagnostic performance of CT, with sensitivity reported as 76-
This work was supported in part by the American 96% [1 5, 1 9-22] and specificity reported as 70-99% [1 5, 20-22]. However, when
Cancer Society JFAA-163 and PDT-326 and the
U.S. Department of Energy DE-FGO2-86ER60418. pathologic inspection is used as a more exacting gold standard, the true sensitivity
I Department of Radiology, Massachusetts Gen-
of contrast-enhanced CT for detection of individual hepatic lesions has recently
eral Hospital and Harvard Medical School, Boston, been shown to be only 38% [23].
MA 02114. Several studies have concluded that unenhanced MR imaging can equal or
2 Present address: Institut f#{252}r
Diagnostische Ra-
exceed the accuracy of contrast-enhanced CT [1 9-21 , 23-25]. The introduction
diologie, Kantonsspital, CH-9007 St. Gallon, Swit-
of superparamagnetic iron oxide as a tissue-specific MR contrast agent for the
zerland. Address reprint requests to C. J. Fretz.
reticuloendothelial system (RES) appears to offer a further improvement in the
3 Department of Radiology, The University of
Chicago, Chicago, IL 60637.
detection of hepatic lesions [26]. In this study, we analyzed clinical data to compare
4 Present address: Nuclear Magnetic Resonance objectively the diagnostic performance of iodine-enhanced CT, unenhanced MR,
Unit, University Hospital, Nuevo Leon, Mexico. and iron oxide-enhanced MR. First, we conducted a receiver-operating-character-
istic (ROC) analysis [27, 28] based on anatomically matched images; second, we
0361 -803X/90/1 554-0763
0 American Roentgen Ray Society compared the number of lesions detected in “complete” CT and MR examinations;
764 FRETZ ET AL. AJR:155, October 1990

and third, we correlated these diagnostic results with quanti- were repeated. The relaxivity, pharmacokinetics, and toxicity of this
tative measurements of lesion-liver contrast-to-noise ratios superparamagnetic iron oxide formulation (AMI-25, Advanced Mag-

(CNR) calculated with standard techniques for MR image netics, Inc., Cambridge, MA) have been described elsewhere [31 , 32]
analysis.

Image Selection
Materials and Methods
The ROC study was based on a total of 478 images representing
Patients 54 anatomic sections (26 normal anatomic sections and 28 abnormal
sections) for which matching enhanced CT, unenhanced MA, and
Ten patients (four men, six women; 41 -65 years of age; mean, 57
enhanced MA images could be identified. Of these 478 images, 260
years) with recent contrast-enhanced CT, unenhanced MR, and iron
were without disease and 21 8 with disease. A lesion was either
oxide-enhanced MR examinations and acceptable proof of the pres-
visible in retrospect with all imaging techniques on the same level or
ence or absence of hepatic lesions were selected for this study. Data not visible on the anatomic section of interest or on the two adjacent
from these patients concerning perfusion and retention imaging for
sections. Sections in which one of the techniques did not have a
lesion detection and characterization have been reported [29, 30].
corresponding image were excluded from the study. This was often
No patients were excluded on the basis of image quality. All 10
the case for the most cephalic and caudal sections. By selecting
patients had a biopsy-proved primary cancer (four colorectal and two
matching anatomic sections instead of “complete” CT and MA ex-
breast adenocarcinorna, and one each had lymphoma, carcinoid, islet
aminations, we could avoid a potential bias against the CT breath-
cell carcinoma, and melanoma). hold imaging technique, which may omit or duplicate one or more
The presence of malignant hepatic lesions in each patient was
anatomic sections because of slice misregistration [33]. In our study,
verified by surgical biopsy (seven), CT or sonographically guided
unenhanced and enhanced MR images were taken at the same level,
percutaneous needle biopsy (two), and autopsy (one) within 1 month
as the position of the center slice was marked on the patient’s skin,
after CT and MR examination. CT or unenhanced MR follow-up
and table position is monitored electronically. No MR or CT images
examinations were performed in five of the patients, and interval
had to be excluded for reasons related to artifacts or poor technical
growth confirmed the presence of individual malignant lesions.
quality.
The interval between CT and MR examinations ranged from 2 to
The proof for presence or absence of individual lesions was based
1 8 days, with a mean of 8 days (CT usually preceding MA) for seven
on intraoperative or pathologic findings in five patients. Twenty-six of
patients. Three patients, one with advanced metastatic disease of
54 anatomic sections were selected from these five patients. Con-
the liver and two patients each with three colonic metastases, had
sensus reading by two radiologists of all MA, sonography, and CT
an interval of 35-64 days between tests, and in these three patients examinations in conjunction with all clinical data and follow-up imaging
there was no demonstrable increase in size or number of lesions.
examinations was used for the remaining five patients, contributing
28 sections to the ROC analysis. A lesion was considered real when
the same defect in a given anatomic section was present on at least
CT Scanning two different types of images, either on CT and MR (whether en-
hanced or not) or on unenhanced and enhanced MR images. Similar
CT examinations were performed on a Technicare 1 440HP unit standards of proof have been used by previous investigators
(four patients: Solon, OH), Somatom DR (one patient; Siemens, Iselin, [19, 25].
NJ), General Electric 9800 (three patients; GE Medical Systems,
Milwaukee, WI), Elscint 2000 (one patient; Hackensack, NJ), or
Technicare 2060 (one patient). Slice thickness was 8 or 10 mm for
Observer Performance Experiments
all examinations. “Contiguous” slices (dependent on patient cooper-
ation) were obtained in nine patients, whereas an intersection gap of Information identifying the patient was obscured on the hard-copy
4 mm was used in one patient. Nine patients received ionic and one films. Individual images were cut from standard multiformat film,
patient nonionic iodinated contrast media (at least 42 g of iodine per mixed in random order, and presented to three radiologists with
patient or approximately 0.6 g of iodine/kg). Injection via peripheral experience in abdominal CT and MR who had not previously seen
vein was performed with a flow of 2 mI/sec by using a mechanical these cases and who did not participate in the consensus reading of
injector in nine patients. CT scanning was initiated 30 sec after the examinations. The readers scored each image for the presence
injection began [22]. A drip infusion was used in one patient who had or absence of focal hepatic lesions and assigned a confidence level
obvious extensive metastatic liver disease. to their observation (1 = definitely or almost definitely absent; 2 =

probably possibly absent; 4

absent; 3 = 5 = possibly present; =

probably present; 6 definitely or almost definitely present).

=

MR Imaging For each imaging method, a binormal ROC curve [28] was fitted
to each observer’s confidence rating data by maximum likelihood
All studies were performed on a 0.6-T (25.1 MHz) superconducting estimation [34]. Diagnostic accuracy of the various imaging methods
imaging system (Technicare, Solon, OH). Baseline unenhanced spin- was determined by calculating the area (Ar) under each reader-
echo (SE) images were acquired with Ti -dependent contrast by using specific binomial ROC curve when it is plotted in the unit square [34,
an SE 260/1 4/1 0 (TR/TE/number averaged) sequence of signals 35]. Composite ROC curves to represent the performance of the
and with intermediate to T2-dependent contrast (SE 500/30/6, SE three readers as a group were calculated by averaging the binormal
1500/40,80/2) sequences. The whole liver could be imaged within 11 parameter values of the individual curves.
sections, each having a thickness of 1 5 mm and an intersection gap Differences between the imaging methods in terms of the mean
of 4 mm. One to two hours after IV infusion of iron oxide particles areas (As) under the ROC curves were analyzed statistically by using
(nine patients received 20 MmoI Fe/kg, and 5 mol Fe/kg were injected Student’s two-tailed t-test for paired data [36]. Differences between
in one patient), the SE 500/30/6 and SE 1 500/40,80/2 sequences ROC curves of individual readers were tested for significance by
AJR:155, October 1990 IRON OXIDE-ENHANCED MR OF HEPATIC METASTASES 765

TABLE 1: Receiver-Operating-Characteristic Estimates for CT and MR Imaging

A Index
Imaging Technique Mean A Index
Reader 1 Reader 2 Reader 3
Contrast-enhanced CT .658 .646 .708 .670 ± .03
Unenhanced MR imaging
SE 260/14 .860 DD .760 .810±07
SE 500/30 .774 .733 .650 .719 ± .06
SE 1500/40 .702 .701 .431 .611 ± .16
SE 1500/80 .737 .741 .642 .707 ± .06
Enhanced MA imaging
SE 500/30 .845’ .898a .701 .81 5 ± .10’
SE 1500/40 .917c .914c .923 .918 ± .oi
SE 1500/80 .977’ .944’ .795 .905 ± .i0
Note.-DD = degenerate data.
.A, index higher (p
.05) than that of contrast-enhanced
< CT.
b Mean A, index higher (p < .05) than that of unenhanced SE 1500/40.
C A1 index higher (p < .01) than that of contrast-enhanced CT.
d Mean A, index higher (p < .005) than that of contrast-enhanced CT.
S Mean A, index higher (p < .05) than that of unenhanced SE 500/30 and unenhanced SE 1500/80.
,A, index higher (p < .001) than that of contrast-enhanced CT.
g Mean A, index higher (p < .05) than those of all unenhanced MR sequences except SE 260/14.

using the CORROC algorithm (for statistical comparison of ROC Results

curves estimated from correlated data sets) [37]. The first of these
two tests for differences between ROC curves accounts for variation The area under the composite ROC curve for the three
within and between readers but ignores case-sample variation, readers (mean A index) was .67 ± .03 for contrast-enhanced
whereas the second test accounts for case-sample variation and CT, .81 ± .07 for the unenhanced SE 260/1 4 MR sequence,
variation within each reader but ignores between-reader variation and .92 ± .01 for the iron oxide-enhanced SE 1500/40
[36]. As Student’s t-test for paired data does not attempt to account sequence (Fig. 1). The enhanced SE 1 500/40 sequence
for case-sample variation, analysis of the significance of differences reached significantly (p < .005, Student’s t-test) greater
between mean A2 values (Table 1) is not affected by our use of
multiple anatomic sections from each patient. Reader-specific p val-
ues obtained by CORROC in this study must be interpreted with
caution, however, because any correlation between readings of dif-
ferent anatomic sections from a given patient in a given technique 0

served to reduce the effective size of the image sample, thereby -..
..-.
potentially causing CORROC to overestimate statistical significance
......._ -.
(i.e., to underestimate p values). . .

Complementary to the ROC experiment, the number of lesions

z
visible in the complete enhanced CT examination and in each unen- 0
hanced and enhanced MR sequence were counted by two other C.)
radiologists, and the size of all lesions was recorded. The inescapable 4
uncertainty of proof for individual hepatic lesions has been discussed U.
(.
/
before [20]. Wilcoxon’s signed-rank test was applied for statistical w /
> ./
analysis with significance assumed at p < .05. I-
05
/
; .
0
3. I
w
Lesion-Liver CNR I.
I-

Lesion-liver CNR was measured quantitatively to compare unen-

hanced and enhanced MR images. In nine patients receiving 20 mol /#{149} - IRON OXIDE.ENHANCED SE 1500/40

Fe/kg, signal intensities of liver, at least one representative tumor,

II uNENHANcED5E260/14
and background noise including ghost artifacts were measured on - - - CCNTRAST#{149}ENHANCEDCT
each image at the same anatomic level to calculate the CNR as the
0.0 . . . . ‘ . .
tumor-to-liver signal difference scaled to the standard deviation of
0.0 05 10
background noise [38, 39]. CNR is an objective measure of pulse
sequence performance that has been shown to correlate with the FALSE POSITIVE FRACTION
detectability of lesions with different MR sequences [26]. Fig. 1.-Ocmposite receiver-operatlng-characteristic (ROC) curves in-
CNR differences between the MR pulse sequences were analyzed dicate relative accuracies with which focal hepatic lesions are detected by
contrast-enhanced CT, unenhanced MR (SE 260/14), and Iron oxide-
statistically with one-factor analysis of variance(ANOVA)for repeated
enhanced MR (SE 1500/40). Plotted data points represent specific ROC
measures and Fisher’s protected least-significance difference (PLSD) points of each reader for Iron oxide-enhanced MR (squares) and contrast-
multicomparison procedure [40]. enhanced CT (diamonds).
766 FRETZ ET AL. AJR:1 55, October 1990

TABLE 2: Number of Lesions Detected with Various Imaging Techniques by Size of Lesion

Diameter of Lesion
Technique Total
<S mm 5-9 mm 10-19 mm 20-29 mm >30 mm
Contrast- 0 (0) 10 (37) 25 (61) 19 (73) 11 (85) 65 (60)
enhanced CT
Unenhanced MR
SE 260/14 0 (0) ii (41) 35 (85) 25 (96) 12 (92) 83 (77)
SE 500/30 0 (0) 9 (33) 31 (76) 25 (96) 12 (92) 77 (71)
SE 1500/40 0 (0) 8 (30) 16 (39) 18 (69) 10 (77) 52 (48)
SE 1500/80 0 (0) 9 (33) 20 (49) 20 (77) 12 (92) 61 (56)
Enhanced MR
SE 500/30 1 (100) 21 (77) 39 (95) 26 (100) ii (85) 98 (91)’
SE 1500/40 1 (100) 24 (89) 40 (98) 26 (100) 13(100) 104 (96)b
SE 1500/80 1 (100) 24 (89) 39 (95) 26 (100) 13 (100) 103 (95)b

Biopsy/autopsy/follow-up 1 (1 00) 27 (1 00) 41 (1 00) 26 (1 00) 1 3 (1 00) 1 08 (100)

Note-Numbers in parentheses are sensitivities in percent.

. Number of lesions detected higher (p < .05) than that detected with unenhanced SE 1500/40.
b Number of lesions detected higher (p < .05) than those detected with all unenhanced MR sequences and contrast-
enhanced CT.

accuracy than contrast-enhanced CT (Table 1). The same After administration of iron oxide, the SE 500/30 and 1500/
enhanced sequence also yielded a higher mean A value 40 sequences far exceeded their preinjection image contrast
(p < .05) than did the unenhanced SE 500/30 and SE 1500/ (Figs. 2 and 3). Lesion-liver CNR measured in images of nine
80 sequences, and there was a trend toward superiority patients receiving 20 moI Fe/kg showed that the enhanced
(p = .08) over the unenhanced SE 1 500/40 sequence. The SE 1 500/40 sequence yielded a significantly (p < .01,
enhanced SE 1 500/80 sequence showed a higher mean A ANOVA) higher CNR value (1 9.5 ± 1 0.2) than did the unen-
value (p < .05) than all unenhanced MR sequences except hanced SE 500/30 and SE 1 500/40,80 sequences (Table 3).
the SE 260/i 4 sequence. In the enhanced SE 1 500/40 sequence, the CNR magnitude
Each reader achieved a greater area (A2 index) under the increased by 50% compared with the best unenhanced pulse
ROC curve with iron oxide-enhanced MR sequences (SE sequence SE 260/14 (-13.3 ± 6.7).
1 500/40 and SE 1 500/80) than with all unenhanced MR
sequences or contrast-enhanced CT (Table 1 ). Testing the
significance of differences between the A indices of enhanced Discussion
MR sequences and contrast-enhanced CT on a reader-by- Pharmaceutical manipulation of tissue relaxation times by
reader basis, two of three readers reached in each compari- administration of magnetic contrast agents can enhance the
son a significantly greater area under the ROC curve with the diagnostic information in MR images [26, 29]. Superparamag-
SE 500/30 sequence (p < .05), with the SE 1500/40 netic iron oxide particles show a tissue-specific biodistribution
sequence (p < .01), and with the SE 1 500/80 sequence to the reticuloendothelial system (RES), where they predom-
(p < .001). inantly increase transverse (T2) relaxation. Non-RES cells lack
Contrast-enhanced CT depicted 65 lesions, 1 7% fewer
than the most sensitive unenhanced MR sequence (SE 260/
14), which showed 83 lesions (Table 2). The iron oxide- TABLE 3: Quantitative MR Image Analysis
enhanced SE 1500/40 sequence showed 1 04 lesions (19%
more than SE 260/1 4 and 36% more than CT). Both the Tumor-Liver
enhanced SE 1 500/40 and the SE 1 500/80 sequences Imaging Technique Contrast-to-Noise Ratio
(mean ± SD)
showed significantly (p < .05, Wilcoxon signed-rank test)
more lesions than all unenhanced MR sequences and Unenhanced MA
SE 260/14 -13.3 ± 6.7a
contrast-enhanced CT. Lesion size ranged from 3 to 50 mm. SE 500/30 -4.7 ± 3.2
Lesions larger than 3 cm were depicted with a high degree of SE 1500/40 3.6 ± 3.7
sensitivity by all imaging methods. For lesions smaller than I SE 1500/80 4.7 ± 5.9
cm in diameter, iron oxide-enhanced sequences detected Enhanced MR
SE 500/30 18.7 ± 6.4b
twice as many as contrast-enhanced CT or the best unen-
SE 1500/40 19.5 ± 10.2’
hanced MR sequence. Table 2 presents a breakdown of the SE 1500/80 15.5 ± Sib
lesions detected and missed with each imaging technique.
Lesion-liver CNR on unenhanced MR images was highest Note-Lesion-liver contrast-to-noise ratio (CNR) calculated in nine patients
receiving 20 1111O1 Fe/kg. An analysis of variance was performed.
for the SE 260/1 4 sequence (Table 3). The SE 500/30 and . CNR higher (p < .01) than that of unenhanced SE 1500/40.
1 500/40,80 sequences showed CNR values close to zero b CNR higher (p < .01) than that of unenhanced SE 500/30 and SE
reflecting tumor-liver isointensity and poor lesion conspicuity. 1500/40,80.
AJR:155, October1990 IRON OXIDE-ENHANCED MR OF HEPATIC METASTASES 767

D E F

Fig. 2.-Metastatic colonic cancer.

A, Unenhanced CT scan. No lesion visible.
B, Iodine-enhanced CT scan. Note good opacification of hepatic vessels. Streaklike artifacts partially obscure left lobe of liver. No lesion visible.
C, SE 260/14 MR image obtained before iron oxide injection shows one lesion (curved arrow) in left lobe of liver.
D and E, SE 500/30 (D) and SE 1500/40 (E) MR images obtained before iron oxide injection. Lesion in left lobe of liver not visible.
F, SE 1500/40 MR image obtained after injection of 20 MmoI Fe/kg of AMI-25. Lesion in left lobe (curved arrow) confirmed; additional lesion (straight
arrow) visible in right lobe.

phagocytic ability, and therefore the relaxation times of tu- of images shows that improved lesion conspicuity and im-
mors remain virtually unchanged after iron oxide administra- proved detection of small lesions (Fig. 2) correspond to the
tion. The resulting loss of signal intensity from liver, with higher A values for the iron oxide-enhanced MR images. Our
tumor unchanged, increases tumor-liver contrast (Figs. 3B- ROC study design may even underestimate the true superi-
3E). ority of iron oxide-enhanced MR imaging, because the dem-
Traditionally, calculation of sensitivity and specificity was onstration of more than one lesion in a given image did not
used to compare new imaging techniques [1 5, 1 9, 25]. How- affect the reader’s scoring (Fig. 3).
ever, there may arise a dilemma in which one technique The number of lesions detected by the enhanced SE 1500/
provides higher sensitivity, but lower specificity, and the rel- 40 sequence was 1 9% and 36% higher than the correspond-
ative capacities of the two techniques cannot be determined ing numbers for the best unenhanced MR sequence (SE 260/
[28]. Instead, ROC analysis provides a more meaningful 14) and contrast-enhanced CT, respectively. The improved
approach to assess diagnostic performance of different tech- contrast between liver and lesion after injection of iron oxide
niques [27, 28]. In this study of iron oxide-enhanced MR allowed a higher detection rate for all enhanced sequences
imaging, the SE 1 500/40 sequence yielded a significantly compared with their unenhanced counterparts. Notably, small
(p < .005) higher mean A value than did contrast-enhanced lesions (5-20 mm in diameter) previously missed on unen-
CT. The three enhanced MR sequences reached higher mean hanced MR or enhanced CT images were frequently visual-
A2 values than did all unenhanced MR sequences. Inspection ized by enhanced MR (Table 2 and Fig. 3). This clinical result
768 FRETZ ET AL. AJR:155, October 1990

Fig. 3.-Metastatic islet cell carcinoma. External biliary drainage with air in biliary tree. Ascites. Portal hypertension with esophagogastric vances.
A, Iodine-enhanced CT scan shows subcapsular hypervascular metastasis in right lobe of liver (arrows).
B and C, SE 500/30 (B) and SE 1500/40 (C) MR images obtained before iron oxide injection. Metastasis is hypointense in B (arrows), not visible in C
because of isointensity with surrounding liver.
D and E, SE 500/30 (D) and SE 1500/40 (E) MR images obtained after injection of 20 Mmol Fe/kg of AMI-25. Liver signal intensity decreased in both
images resulting in better delineation of 2.5-cm subcapsular metastasis. In addition, two small (<5 mm) metastases (arrows) are only visible on enhanced
MR images because of significantly increased lesion-tumor contrast.

confirms animal experiments showing improved detection of quence achieving the highest CNR (enhanced SE 100/40)
millimeter-sized lesions by iron oxide-enhanced MR imaging found the greatest number of lesions (Table 2) and yielded
[26]. the highest ROC indexes (Table 1).
Lesion-liver CNR increased significantly (p < .01) for MR The technique of contrast administration (drip infusion) used
sequences after IV administration of iron oxide (Table 3). The in one of our patients did not contribute to the poor perform-
increased CNR improved lesion recognition. The superiority ance of CT, because a separate analysis excluding this patient
of MR imaging over CT is apparently due to the higher tumor- showed the same results (in terms of the number of lesions
liver contrast achieved by MR imaging. Unfortunately, com- detected). Claims have been made that technical variations in
parable quantitative analysis of contrast and noise on CT the administration of iodinated contrast agents may allow
scans is not possible, and for comparisons we have to rely detection of up to 1 5% more lesions [1 7] or hide up to 13%
on our subjective visual impression. Indeed, inspection of of lesions [41 1. However, these two studies did not test the
images confirms that increased tumor-liver contrast allows statistical significance of their observations. Our results show
the detection of additional lesions, particularly smaller ones 36% more lesions detected with enhanced MR sequences
(Figs. 2 and 3). Animal experiments have shown that greater than with CT. This increase is twofold to threefold greater
image CNR lowers the threshold size at which metastases than differences among various CT enhancement methods
are detected [26]. Therefore, it is no surprise that the se- [16, 17, 41].
AJA:155, October1990 IRON OXIDE-ENHANCED MR OF HEPATIC METASTASES 769

The conventional SE pulse sequences used in this study cology. Philadelphia: Lippincott, 1982:1553-1563
4. Wagner JS, Adson MA, van Heerden JA, Adson MH, llstrup DM. The
did not suppress or refocus signalfrom flowing blood, creating natural history of hepatic metastases from cob-rectal cancer: a comparison
some difficulty in distinguishing peripheral hepatic vessels with resective treatment. Ann Surg 1984:199:502-508
from small lesions. This was particularly true for iron oxide- 5. Foster JH. Survival after liver resection for secondary tumors. Am J Surg
enhanced MR imaging, because enhanced MR images dis- 1978:135:389-394
6. Foster JH, Lundy J. Liver metastases. In: Ravitch MM, ed. Current prob-
played more peripheral vascular structures than did contrast-
lems in surgery. Chicago: Year Book Medical, 1981 : 161 -202
enhanced CT scans. Furthermore, small vessels and lesions 7. Cady B, McDermott WV. Major hepatic resection for metachronous me-
exhibit similar signal intensities on iron oxide-enhanced MR tastases from colon cancer. Ann Surg 1985:201:204-209
images, whereas they often (but not always) show different 8. Taylor I. Colorectal liver metastases-to treat or not to treat? Br J Surg
X-ray attenuation on contrast-enhanced CT scans. This pos- 1985:72:511-16
9. Brown DA, Pommier RF, Woltenng EA, Fletcher WS. Nonanatomic hepatic
sibility of distinguishing vessels from lesions on contrast-
resection for secondary hepatic tumors with special reference to hemo-
enhanced CT may even have favored CT. In the future, this static technique. Arch Surg 1988:123: 1063-1 066
difficulty may be overcome with the implementation of rapid 1 0. Adson MA. Cannon lecture: hepatic metastases in perspective. AJR
MR imaging during the injection of iron oxide [30] or the use 1983;140:695-700
1 1 . Malt RA. Current concepts: surgery for hepatic neoplasms. N EngI J Med
of pulse sequences designed to identify the signal of flowing
1985:313: 1591-1 596
blood [42]. Nevertheless, even without any special pulse- 12. Hughes KS, Simon R, Songhorabodi S. et al. Resection of the liver for
sequence techniques or flow compensation, enhanced MR colorectal carcinoma metastases: a multi-institutional study of indications
proved superior to enhanced CT in this study. for resection. Registry of hepatic metastases. Surgery 1988:103:278-288
Several groups have shown that unenhanced MR is com- 1 3. Alderson P0, Adams DF, McNeil BJ, et al. Computed tomography, ultra-
sound, and scintigraphy of the liver in patients with colon or breast
petitive with contrast-enhanced CT for assessment of hepatic
carcinoma: a prospective comparison. Radiology 1983; 149 : 225-230
tumors [1 9-25, 43]. Our results indicate that the diagnostic 1 4. Freeny PC, Marks WM. Computed tomographic artenography of the liver.
performance of hepatic MR imaging can be further improved Radiology 1983:148:193-197
by administration of iron oxide. Reviewing the data justifying 1 5. Sugarbaker PH, Vermess M, Doppman JL, Mifler DL, Simon R. Improved
detection of focal hepatic lesions with computerized tomographic exami-
contrast-enhanced liver CT, we see a greater impact of con-
nation of the liver using ethiodized oil emulsion (EOE-1 3) liver contrast.
trast media on liver MA, even at this early stage of develop- Cancer 1984:4:1489-1495
ment. In a recent study of iron oxide-enhanced MR at 1 .5 T, 16. Bernardino ME, Erwin BC, Steinberg HV, Baumgartner BR, Torres WE.
Marchal et al. [44] showed similar gains in terms of CNR and Gedgaudas-McClees AK. Delayed hepatic CT scanning: increased confi-
improved performance of standard SE (moderate TA) pulse dence and improved detection of hepatic metastases. Radiology 1986:
159:71-74
sequences when enhanced by iron oxide. Indeed, their en-
17. Paushter DM, Zeman AK, Scheibler ML, Choyke PL, Jafte MH, Clark LA.
hanced SE 800/60 sequence showed more lesions than their CT evaluation of suspected hepatic metastases: comparison of techniques
enhanced SE 2200/70 sequence in a much shorter acquisition for IV contrast enhancement. AJR 1989:152:267-271
time. However, both enhanced sequences visualized a slightly 1 8. Bruneton JN, Falew#{233}eMN, Fran#{231}ois
E, et al. Liver, spleen, and vessels:
preliminary clinical results of CT with perfluorooctylbromide. Radiology
smaller number of lesions than did their standard unenhanced
1989:170:179-183
sequence for clinical evaluation of liver lesions (SE 2200/70 19. Reinig JW, Dwyer AJ, Miller DL, et al. Liver metastasis detection: compar-
with gradient nulling motion compensation), and they con- ative sensitivities of MR imaging and CT scanning. Radiology 1987:162:
cluded that the benefit of iron oxide enhancement was only 43-47
marginal. Our results at a midfield system clearly differ from 20. Stark DD, Wittenberg J. Butch RJ, Ferrucci JT. Hepatic metastases:
controlled comparison of detection with MR imaging and CT. Radiology
those of Marchal et al. Iron oxide-enhanced MA significantly
1987:165:399-406
improved detection of liver metastases compared not only 21 . Chezmar JL, Rumancik WM, Megibow AJ, Hulnick DH, Nelson RC, Ber-
with sequences with known low hepatic tumor conspicuity nardino ME. Liver and abdominal screening in patients with cancer: CT
(SE 500/30 and SE 1500/40,80) [38] but also with our stand- versus MR imaging. Radiology 1988:168:43-47
ard unenhanced sequence for liver tumor screening (short 22. Ferrucci JT, Freeny PC, Stark DD, et al. Advances in hepatobiliary radiol-
ogy. Radiology 1988:168:319-338
TA/short TE). With careful pulse-sequence optimization, the 23. Heiken JP, Weyman PJ, Lee JK, et al. Detection of focal hepatic masses:
performance of MA with respect to lesion-liver contrast or prospective evaluation with CT, delayed CT. CT during arterial portogra-
examination time will increase even further [45]. The apparent phy, and MR imaging. Radiology 1989:171:47-51
superiority of enhanced MR imaging over iodine-enhanced 24. Curati WL, Halevy A, Gibson RN, Carr DH, Blumgart LH. tJtrasound, CT,
and MRI comparison in primary and secondary tumors of the liver. Gas-
CT of the liver now justifies prospective, multicenter trials to
trointest Radiol 1988:13: 123-1 28
establish the role of iron oxide-enhanced MA in screening 25. Reinig JW, Dwyer AJ, Miller DL, Frank JA, Adams GW, Chang AE. Liver
patients at risk for cancer. metastases: detection with MR imaging at 0.5 and 1 .5 T. Radiology
1989:170: 149-1 53
26. Tsang YM, Stark DD, Chen MC, Weissleder A. Wittenberg J. Ferrucci JT.
REFERENCES Hepatic micrometastases in the rat: ferrite-enhanced MR imaging. Radiol-
ogy 1988:167:21-24
1 . Gilbert HA, Kagan AR. Metastases: incidence, detection, and evaluation 27. Swets JA. Measuring the accuracy of diagnostic systems. Science
without histologic confirmation. In: Weiss L, ed. Fundamental aspects of 1988:240:1285-1293
metastases. Amsterdam: North Holland, 1976:385-405 28. Metz CE. ROC methodology in radiologic imaging. Invest Radiol 1986;
2. Ozarda A, Pickren J. The topographic distribution of liver metastases: its 21 :720-733
relation to surgical and isotope diagnosis. J Nucl Med 1962:3:149-152 29. Stark DD, Weissleder R, Elizondo G, et al. Superparamagnetic iron oxide:
3. Foster JH. Treatment of metastatic cancer to liver. In: DeVita VT Jr, clinical application as a contrast agent for MR imaging of the liver. Radiol-
Hellmann 5, Rosenberg SA, eds. Cancer: principles and practice of on- ogy 1988:168:297-301
770 FAETZ ET AL. AJR:155, October 1990

30. Hahn PF, Stark DD, Weissleder A, Elizondo G, Saini 5, Ferrucci JT. Clinical analysis of pulse sequence performance in MA imaging. Radiology
application of superparamagnetic iron oxide to MR imaging of tissue 1986;159:365-370
perfusion in vascular liver tumors. Radiology 1990;174:361-366 39. Stark DD, Hendrick RE, Hahn PF, Ferrucci JT. Motion artifact reduction
31 . Weissleder A, Stark DD, Engelstad BL, et al. Superparamagnetic iron by fast spin echo imaging. Radiology 1987;164: 183-1 92
oxide: pharmacokinetics and toxicity. AJA 1989;152:167-173 40. Armitage P. Berry G. Statistical methods in medical research, 2nd ed.
32. Josephson L, Lewis J, Jacobs P, Hahn PF, Stark DD. The effects of iron Oxford, England: Blackwell Scientific, 1987
oxides on proton relaxivity. Magn Reson Imaging 1988;6:647-653 41 . Moss AA, Schrumpf J, Schnyder P, Korobkin M, Shimshak AR. Computed
33. Foley WD. Commentary: dynamic hepatic CT scanning. AJR 1989;152: tomography of focal hepatic lesions: a blind clinical evaluation of the effect
272-274 of contrast enhancement. Radiology 1979;131 :427-430
34. Swets JA, Pickett AM. Evaluation of diagnostic systems: methods from 42. Felmlee JP, Ehman AL. Spatial presaturation: a method for suppressing
signal detection theory. New York: Academic Press, 1982 flow artifacts and improving depiction of vascular anatomy in MA imaging.
35. Hanley JA, McNeil BJ. The meaning and use of the area under a receiver Radiology 1987;164:559-564
operating characteristic (ROC) curve. Radiology 1982;143:29-36 43. Moss AA. Editorial: imaging of colorectal carcinoma. Radiology 1989;170:
36. Metz CE. Some practical issues of experimental design and data analysis 308-310
in radiological ROC studies. Invest Radiol 1989;24:234-245 44. Marchal G, Van Hecke P, Demaerel P, et al. Detection of liver metastases
37. Metz CE, Wang P-L, Kronman HB. A new approach for testing the with superparamagnetic iron oxide in 1 5 patients: results of MA imaging
significance of differences between ROC curves measured from correlated at 1.5T.AJR 1989;152:771-775
data. In: Deconinck F, ed. Information processing in medicalimaging. The 45. Fretz CJ, Elizondo G, Weissleder A, Hahn PF, Stark DD, Ferrucci JT.
Hague: Nijhoff, 1984:432-445 Superparamagneticiron oxide enhanced MRI: pulse sequence optimization
38. Stark DD, Wittenberg J, Edelman A, et al. Detection of hepatic metastases: for detection of liver cancer. Radiology 1989;172:393-397