Eric
Outwater,
Hepatic of MR
MD
John
#{149}
E. Tomaszewski,
Colorectal Imaging and
Results of magnetic resonance (MR) imaging examinations for 76 patients with proved colorectal metastases to the liver were retrospectively reviewed. Signal intensity and morphologic patterns of 157 lesions were scored. The morphologic patterns were correlated with the histologic material obtained at segmental hepatic resections or excisional biopsies in 33 patients. In 154 lesions (98%) colorectal metastases to the liver were found to be hyperintense on MR images acquired with a long repetition time/echo time; in 77 lesions (49%) central low-signal-intensity changes were seen. Virtually all larger lesions demonstrated areas of low signal intensity within the himor. These areas were found to reflect histologic findings of coagulafive necrosis and desmoplasia within the tumor. Peripheral hyperintense halos around central hypointense areas encompassed the growing himor margin and variable degrees of cell necrosis. Hypointense peripheral rims, which correspond to abnormalities of surrounding hepatic parenchyma, were seen in 40 lesions (25%). This sign should not be assumed to represent the fibrous pseudocapsule of primary hepatic malignancy. No correlation between tumor grade (differentiation) and tumor morphology was observed. Index terms: Colon, neoplasms, 75.321 #{149}Liver neoplasms, MR studies, 76.1214 #{149}Liver neoplasms, secondary, 76.332 #{149}Rectum, neoplasms, 757.321
Radiology
‘From
1991; 180:327-332
the Departments
of Radiology
(E.O.,
H.Y.K.), Pathology (JET.), and Surgery (J.M.D.), Hospital of the University of Pennsylvania, 3400 Spruce St, Philadelphia, PA 19104. Received January 17,1991; revision requested February 12; revision received March 22; accepted April 1. Supported in part by NIH grant no. 1-ROlCA47354-0l. Address reprint requests to H.Y.K. ORSNA,
1991
MD
John
M. Daly,
#{149}
Metastases: Pathologic
MD
#{149} Herbert
importance of abdominal magnetic resonance (MR) imaging in the evaluation of liver metastatic disease is increasing, owing to its favorable sensitivity and specificity cornpared with those achieved with computed tomography (CT). To date, studies have focused predominantly on the efficacy of MR imaging in the detection of metastases, the optimal pulse sequences for their detection, and the differentiation of benign and malignant lesions based on morphologic patterns and signal intensity (15). As metastases in these studies tend to be treated in the aggregate, an understanding of the spectrum of MR appearances of metastases of any one histologic type is lacking. The variable MR imaging appearance of rnetastases to the liver was observed early in dinical studies of hepatic lesions. This variability may be seen even within one patient with one primary tumor (5). Although have
been
developed
the classification of MR pafterns hepatic mass lesions as an aid in tinguishing benign lesions from lignant (2,5), the microanatomic lates to the diverse metastases are not This retrospective
of dismacorre-
patterns of well characterized. MR study of he-
Patient
AND
mor and proved hepatic metastases. Five patients were exduded from analysis; three patients did not have images available for review, one patient had findings markedly obscured by recent surgical changes, and one patient with normal findings on MR images was subsequently shown to have metastasis at biopsy 3 months after the MR imaging examinalion. The remaining 76 patients comprised the study group. Medical records for each patient were reviewed for clinical data. The mean age of the patients was 62 years (range, 19-84 years);
46 were
men,
and
30 were
women.
The primary tumor was adenocarcinoma of the colon in 56 patients and rectal cardnoma in 20. Twenty-one patients had received systemic chemotherapy, usually fluorouracil, months to years prior to the MR imaging examination. Three of these patients received chemotherapy via an intraarterial infusion pump in the hepatic artery. Forty-nine patients had received no treatment for metastatic disease before the MR examination. In four cases, the previous treatment history could not be obtained.
for
patic metastases was performed with two goals in mind. First, we wished to establish the frequency and type of MR patterns in a single histologic type of metastasis, namely adenocarcinoma of the colon and rectum. Second, we have attempted to establish the microanatomic basis for these signal patterns.
PATIENTS
MD
Correlation Appearance’
T
schemata
Y. Kressel,
METHODS
Population
A computer search of records of patients who had undergone MR examinations on a 1.5-T system at this facility from January 1986 to October 1990 disdosed 81 patients with a history of a colorectal primary hi-
MR
Imaging
MR imaging system (Signa; waukee) with was an indusive tion extending pulse
sequences
was
performed
on a 1.5-T
GE Medical Systems, the body coil. Because retrospective
been
this
investiga-
back to 1986, a variety had
Mil-
of
employed.
Studies consisted of spin-echo images obtained in the axial plane with short repetition time (TR)/echo time (TE), long TR/ short TE, and long TRiTE sequences. Additional sequences were performed in many patients but were not used in the analysis of the studies. For the short TR[FE series, TRs ranged from 300 to 800 msec; 88% of the studies were acquired with TRs of either 400 or 600 msec. TEs varied from 25 to 12 msec. Images were obtained with two to four acquisitions and a 256 x 128 matrix, with a section thickness of 10 mm. The long TR images used a TR of 2,500
Abbreviations: tion time.
TE
=
echo time, TR
=
repeti-
327
msec
in 69 of the 76 patients;
nations
seven
exami-
were
performed with longer TRs. Symmetric TEs of 40 and 80 msec were used in 68 of the 76 patients; asymmetric echoes of 20-30 and 80 msec were used in the remaining patients. The section thickness was 5-10 mm, and an acquisition trix of 256 x 128, with two acquisitions,
ma-
area) for correlation of the pathologic and MR image features. Because this was a retrospective study, precise localization of the tissue block to the imaged tumor was not possible; however, in each case the tumor edge/liver interface could be exammed. Each of these histologic specimens was retrospectively reviewed by one pa-
was used in most cases. Respiratory cornpensation by means of reordering of phase-encoding steps and gradient-mornent nulling was generally employed. In eight patients, hepatic resection specimens were imaged with MR prior to fixation. A 10-16-cm field of view, 5-mm section thickness, and spin-echo sequences of 400-600/20 and 2,500/40, 80 were used to image the specimens. One image per patient was retrospeclively analyzed by two of the authors. If
O.T.) for specific features in the and surrounding liver. These induded the relative amounts of desmoplastic reaction, tumor necrosis, mucin accumulation within the tumor, and tumor
patients
In the 76 patients, 147 individual tumor nodules were identified on the MR images, and 10 patients had metastases that were too numerous to count or were confluent. The latter
had
more
than
one
image
avail-
able, the study performed dosest in time to the liver biopsy was chosen for analysis. Every lesion in the liver was examined for size, morphology, and signal intensity. Lesions that became progressively more intense
relative
to liver
on
heavily
segmental when
hepatic
lesion
or area
of tumor.
such as adenopathy edema were noted
or
present.
The size of each tumor mass was measured at its greatest diameter in the axial plane, with the accompanying perpendicular diameter. The size is expressed as the product of these diameters in square centimeters.
Pathologic
Review
The presence
of metastatic
disease
these patients was established biopsy in 67 patients, needle cytology in six patients, and
the hepatic
metastases
biopsy of the hepatic patients. Thirty-three
tumor
nodules)
tectomy,
wedge
resection,
of
aspiration palpation
metastases patients
underwent
in
by means
at surgery
of
without in three (with 36
partial
hepa-
or excisional
biopsy of their metastasis to yield enough tissue (approximately 1-4-cm2 section 328
Radiology
#{149}
tumor
grade.
RESULTS MR Imaging Metastases
lesions
T2-
weighted sequences (2,500/120-160) and met standard morphologic criteria for cysts or hemangiomas were exduded from analysis. Signal intensity of the components of the individual lesions was subjectively scored on a seven-point scale (1 = very high signal intensity, equal to that of cerebrospinal flu. on a long TRiTE sequence or fat on a TI-weighted sequence; 2 = hypetinten compared with liver; 3 = slightly hyperint.nse compared with liver; 4 = isointense with liver or not seen; 5 = slightly hypointense compared with liver; 6 = hypointense compared with liver; 7 = very low signal intensity, equal to that of simple fluid on a short TRIFE sequence). Lesions that were confluent or too numerous to count were treated for the purposes of scoring as a solitary lesion having the signal characteristics of the largest Ancillary findings
thologist
were
Characteristics
treated
of
as solitary
or not
seen,
lesions
as opposed
to
three of 157 (2%) on the long TR/TE images (Table 1). Part of the failure to visualize these lesions on the short TRIFE images, however, may be due to the thicker sections used in some cases for this sequence, as well as the TEs of up to 25 msec. Intermediateweighted sequences (short TFllong TR images) rendered more lesions isointense or slightly hyperintense to liver. Hence, this sequence appeared to be inferior to that of long TRTFE images for the detection of metastatic deposits.
Of the 157 lesions, strated central areas lower signal intensity higher-intensity
tumor
77 (49%) that were relative edge
demonof to the on
inhomogeneity was found patients with a history of syschemotherapy (44% preva-
temic lence). In contrast, focal areas of very high signal intensity on the T2-weighted images correlated with areas of very low signal intensity on the TIweighted images. These signal characteristics are indicative of fluid, and liquefactive necrosis appeared to account for the high signal intensity (Fig 1). This central high intensity occurred in 24 of 157 lesions (15%) and was seen only in larger lesions (Fig 2). In 21 of 24 cases (88%), areas of very high intensity within the tumor were surrounded
for the purposes of analysis, yielding a total of 157. As expected, the predominant signal intensity of the himors was moderately low relative to normal liver on short TRITE sequences and moderately high on long TR/TE images. On short TRiTE images, 25 of 157 lesions (16%) were isointense
intensity in those
the
T2-weighted images (Fig 1). These areas varied from small foci to confluent central areas of low signal intensity surrounded by a rim of higher signal intensity. Large lesions were more likely to demonstrate inhomogeneous signal intensity than smaller lesions (Fig 2), although even some small lesions showed centers of low signal intensity. No correlate to these areas on the TI-weighted images could be found, that is, both the lowintensity center and the high-intensity rim showed uniformly low intensity on the short TRIFE images. No higher frequency of central signal-
by
areas
of low
intensity.
Only four lesions contained foci of high intensity on short TRIFE images. These areas were also of high intensity on the T2-weighted images, consistent with intratumoral hemorrhage or highly proteinaceous fluid. In 40 lesions (25%), a thin rim of homogeneous low signal intensity surrounded the high-signal-intensity tumor on long TR/short TE and long TRTFE images (Fig 3). The rim varied from 2 to 8 mm thick and increased the contrast and distinctness of the border between the tumor and the surrounding liver. In seven lesions, a low-signal-intensity rim appeared on the Ti-weighted observed only
isointense
images. in lesions
or nearly
This was showing
signal images in contrast to lesions with a low-signal-intensity rim (Figs 4, 5). Similar to lesions with the low-signal-intensity characteristics
centers,
on
these
isointense
short
TR/TE
low-intensity
rims
pre-
dominated also seen sions. slightly
in larger lesions but were around some smaller leIn two patients a faint rim of high
intensity
surrounded
the
low-intensity tumor on short TR/FE images. Long TR/TE images in five patients demonstrated discrete wedge-shaped areas of slightly high signal intensity corresponding to vascular territories. These areas were associated with centrally located tumors in a position to compress or ocdude portal vessels or segmental bile ducts.
Histologic Examination
Findings of the
histologic
sec-
tions from 33 patients demonstrated a generally stereotypic histologic pattern. Viable tumor cells tended to cluster within several millimeters of the periphery of the nodule and were admixed with small areas of mucin and cell necrosis. More centrally, large August1991
Table 1 Predominant Metastases
Signal Intensity of 157 Related to Pulse Sequence Pulse Sequence
Signal Intensity Rating
5,
,
r”v: ,
L
.p
0
Short TR/TE
Long TR/ Short TE
1
0
0
7
2 3 4 5 6 7
0 0 25 43 88 1
53 70 32 1 1 0
119 27 3 0 1 0
100 95 90 85 80 75 70
.c
w .
.2
ll 0
C, C-
65 60 55 50 45 40 35 30 25 20 15 10
I
H1
NR 3
5
Lesion
Figure d.
C.
Figure
1. Lowand high-internal-signal-intensity changes. (a) Long TRITE (2,500/80) shows hyperintense edge of tumor, confluent area of low signal intensity internally, punctate region of high signal intensity centrally. (b) Axial (2,500/80) image obtained surface coil of left lobectomy specimen (c) shows the well-demarcated hypenntense rim corresponding to the histologic finding of advancing cellular tumor rim (arrows
(c) Large-mount
section
of right half of tumor
(trichrome
stain).
Areas
of low signal
image and with a tumor
in c). intensity
2.
Signal
10
30
Size
intensity
100
I 200
300
TNTC
(cm
patterns
related
to
lesion size. Lesion size is expressed as the product of maximum perpendicular diameters in the axial plane. Dotted bars represent lesions with foci of central low signal intensity. Patterned bars represent lesions with low-signal-intensity rims. Black bars repre-
sent lesions sity. TNTC
with =
central
too
high
numerous
signal
inten-
to count.
on MR stained represent
stain
images correspond to the areas of coagulative necrosis (arrowheads), which are dark red, and fibrosis, which are stained blue. High-signal-intensity foci in a and b accumulations of cell debris (*) in the center of the tumor. (d) Use of prussian blue fails to demonstrate any hemosiderin deposits to account for the low-intensity areas. L = liver.
Long TR/TE
areas
of coagulative
plasia, and, accumulation Viable
tumor
mediate were the
cells
edge usually
All tumor and
to represent
desmo-
from
tumor
sparse
compared and
nodules
the
without
low-intensity
long
TRiTE
nent from
that the
and
images.
was tumor
centers
The
tissue
on
compo-
most prevalent away edge was determined
correlated
with
the
MR
imaging
signal pattern. These results are summarized in Table 2. Although other central tissue components were usually present, focal and confluent centers of low intensity within the tumor on the long TR/TE images appeared Volume
180
Number
#{149}
2
cellular
from no
histologic feature absolutely discriminated between tumor nodules with and
more rim.
areas
At histologic
some
Therefore,
of more
marked
near
the
tumor
examinations,
low-
signal-intensity rims around nodules on the T2-weighted
debris.
desmoplasia,
production.
areas
desmoplastic reaction and coagulative necrosis. These features blended with
with cell
displayed
necrosis,
im-
nodule
of fibrosis
of cell mucin
away
of the
amount
degree
necrosis,
to varying degrees, mucin predominated (Fig 1).
12 of 33 patients
represent hepatic
tumor images
were
pathologic parenchyma
found
changes adjacent
to
in the to the
tumor.
Compression
of hepatic
paren-
chyma atrophy,
and sinusoids, and fibrosis
hepatocellular occurred most
frequently (Fig 3), with foci of hemorrhage, inflammation, and congested sinusoids
sometimes
demonstrates peritumoral
encountered. ity
as determined
specimens-that from normal
one fibrotic
The
seen.
of the
Figure
more
reactions
3d
marked that
zone
of abnormal-
from
the
histologic
is, the transition liver to tumor edge-
ranged
from
0.3 to 8.0 mm
wider in lesions signal-intensity (mean, 2.7 mm) nodules without images There
(mean, was no
and
was
demonstrating a lowrim on MR images than in those tumor visible rims on MR 0.7 mm; P < .005). histologic
evidence
of
extracellular fluid accumulations around the tumor to indicate the presence of vasogenic edema that could contribute to the signal intensity pattern as seen on MR images. No
histologic
the
faint
correlate
was
found
high-signal-intensity
on Ti-weighted seen around
to
rims
images, which were metastases. Iron of tumors from five patients three
staining was performed and dence of significant in the tumors. were observed mens.
we
Enlarged fled the
in nine presence
revealed no eviiron deposition
No calcium on histologic
lymph
nodes
deposits speci-
were
identi-
patients. In five of these, of lymphatic metastasis
was confirmed at biopsy. tients, lymphadenopathy
In two pawith lipoRadiology
329
#{149}
Figure
3.
Low-signal-intensity
rim
on
long
TR/I’E images. (a) Long TRITE (2,500/80) image demonstrates a metastasis in the right lobe of the liver with a 2-3-mm rim of low signal intensity (arrows). (b) The rim of compressed hepatic parenchyma (between arrowheads) adjacent to the tumor edge corresponds to the low-intensity rim seen in a.
(Hematoxylin-eosin
stain;
original
magnifica-
tion, x2.) (c) A lesion in another patient shows a similar low-intensity rim on this spin-echo (2,500/80) image, representing the histologic finding (d) of a rim of fibrous tissue (arrowheads) between hepatic parenchyma (L) and tumor (fl. (Hematoxylin-eosin stain; original magnification, x2.)
granulomas without found at biopsy. The
metastasis other two
was pa-
tients did not have surgical confirmation. Two patients had periportal lymph node metastases that were not seen
on
MR
moderately
slides diswere in eight
to poorly
differ-
entiated in 21 patients, and poorly differentiated in four. No definite terns on MR images corresponded these
b.
images.
Review of the histologic closed that the carcinomas moderately differentiated patients,
a.
histologic
grades,
as signal
patto inho-
mogeneity was similar for these groups. Similarly, no definite correlation emerged of treatment
between a prior with chemotherapy
the histologic aging pattern. bid high
features, signal
history and
findings or the MR imCarcinomas with colhowever, intensity
on
images
in three
of four
paucity
of focal
low-intensity
showed very T2-weighted
cases,
with
a
areas ‘
(Fig 5).
DISCUSSION ‘.34’
Colon and rectal adenocarcinomas are among the most frequent primary tumors to metastasize to the liver and are the most amenable to surgical resection. For MR imaging to play an important
role
in the
management
these cases it must demonstrate number of metastases present
of
the with
t..
high sensitivity, enable distinction between benign and malignant lesions with high specificity, and help establish the extent of each lesion to
determine ing
its resectability.
at 1.5 T, the
long
TRITE
For imagimage
has
the highest sensitivity for hepatic metastases (3) and provides the critical morphologic and signal intensity information to help distinguish benign from malignant lesions (5). With these sequences, however, colorectal metastases manifest logic and signal
Accordingly, 330
#{149} Radiology
a variety intensity
a knowledge
of morphopatterns.
of the
4
b.
a.
Figure
4. Low-signal-intensity shows isointense tumor mass fibrosis and hepatic compression eosin stain).
rim
on short TRITE with thin hypointense adjacent to tumor
range of MR imaging patterns and their meaning with long TRIfE sequences is important. Review of the MR findings in the 157
metastases
in this
series
supports
the findings of others that long TRITE images are superior to short TRITE or
image. (a) Ti-weighted (400/15) image rim (arrows). (b) Section demonstrates edge (between arrows) (hematoxylin-
long
TR’ short
TE images
in demon-
strating metastases at 1.5 T (3). Approximately one-quarter and one-
third, respectively, of the individual lesions were isointense on the latter sequences. Only one-half, however, were relatively homogeneous in inAugust
1991
Table 2 Correlation of Histologic Findings Long TRITE MR Imaging Findings Tumor Nodules from 33 Patients Internal
and in
Signal
Intensity Tissue
Component
Coagulative (n=16)
High
Low
necrosis
Fibrosis (i = 9) Tumor cefls
(n=4) Mucin/cell (n=3)
7
9
0
9
4
0
2
1
1
0
debris
Liquefactive
necrosis
(n=1)
with peritumoral edema. We found no examples of low-signal-intensity -
I
-
.
nodules
without
Second,
the
a hyperintense
size
of the
rim.
metastases
found at surgery corresponded size of the hyperintense mass weighted
MR
pointense was often
central area alone, very small relative
metastases
on
sharply tumor
from
5.
Colloid carcinoma. (a) Short TRITE (400/20) image shows a thin hypointense rim surrounding an isointense tumor with a hypointense center. (b) Histologic specthe edge of the tumor (T) shows a rim of compressed hepatic parenchyma associ-
(arrowheads) imen
from
ated with obliteration
of mild steatosis
present
stain). (c) Long TRITE tensity corresponding histologic section (d).
(2,500/80) image to accumulations
of the same
tensity
on
degree
of relative
sity
was
many
long
of lower
TR/TE
images.
central
areas foci
intensity
a thin areas
areas
were
lower-intensity
of signal change were correlated
metastasis;
to suggest these patwith hypoin-
virtually
small,
cen-
surrounded rim. The intensity
surrounded areas.
region
surrounded
from
to large
hyperintense of high signal
usually types mor
ranging
of nodules;
a confluent
by the hyperintense rim the “halo” sign. However, terns form a continuum, tral
by central (fluid)
by these Both
of these
180
demonstrates
the marked
(*) and cell debris within
hyperin-
the tumor
on the
#{149} Number
if the
images. The clinical significance this is that this peripheral zone
growth.
It has been suggested that sign on T2-weighted images sents ule
a low-intensity surrounded
the halo repre-
metastatic by
edema
nod-
within
the
liver, based on clinical studies of metastases of various primaries (5,6). We found no evidence to suggest that this mechanism is responsible for the signal intensity patterns of colorectal metastases.
This
conclusion
is based
metastases a low-intensity
sity rind, of nodules
that
includes center
lesions and
with
high-inten-
rather than a distinct group wholly of low intensity
a
area
We conclude, therefore, majority of cases, concentric edema does not contribute
to tumor
all of the
2
liver,
to occur
significantly
on our findings that, first, there is a spectrum of appearances of colorectal
largest
surrounding
that these internal changes in signal intensity represent a spectrum of internal histologic changes related
within the tuto the size of the
tumors showed some degree of relative high- or low-signal-intensity changes within the tumor. We sugVolume
in the liver (L) (hematoxylin-eosin
lesion
gest
Some
hypointen-
by 49%
had
signal
of mucin
elsewhere
images
hyperintense of
hyperintensity actually represents hepatic parenchymal edema. Lastly, the depth of histologic abnormality in the surrounding hepatic parenchyma around the nodules that were exammed histologically was insufficient to account for the peripheral hyperintensity seen in most of those nodules demonstrating a low-signal-intensity center. in the hepatic
displayed
of these
tense
..-
the
hy-
which to the
region of hylow-signalaround some
the
unlikely
to the T2-
to the
T2-weighted
demarcated
finding
Figure
not
size of the surrounding perintensity. Third, the intensity rims observed
*
/
images,
on
signal tastases
to the
intensity on long
perintensity represent
ring
that,
patterns
of
of colorectal meTRITE spin-echo
should be assumed tumor for the purposes
surgical
planning
and
tumor
MR of of hy-
to of
volumet-
nc analysis. This
is in contrast
shaped
intensity generally the tumor
areas
to the
of slightly
on long extended toward
TRITE
wedge-
high
signal
images
peripherally the hepatic
that from capsule.
Itai et al (7) noted the former finding in six of 73 cases of primary and secondary malignant hepatic tumors and showed struction Rummeny
that intrahepatic portal obmay account for the finding. et al (8) suggested that
Radiology
#{149} 331
lymphatic obstruction may play a role and that the finding was not associated with metastases. We observed this pattern, however, in five patients. Thus, although uncommon, it can occur in association with colorectal metastases. Muramatsu et al (9) have classified colorectal metastases in the liver into three histologic patterns: those demonstrating a completely peripheral distribution of tumor cells and a fibrotic center; those with a predominantly peripheral distribution of himor cells with mixed desmoplasia and carcinoma in the center; and those with mixed carcinoma cells, coagulative necrosis, and fibrosis throughout the mass. Although we found these distinctions somewhat arbitrary, as virtually all of the histologic specimens we reviewed had a predominantly peripheral distribution of carcinoma and very few completely lacked carcinoma cells centrally, this descriptive classification does serve to generally characterize the internal structure of these lesions accurately. Internal desmoplastic reaction is often a dominant feature of these lesions and may contribute to the central areas of low signal intensity seen in some tumors. We found that these areas usually contain a considerable amount of necrotic cellular debris interspersed with the fibrosis, although they have been shown to be perfused on late-enhanced CT scans (9). In the nodules from the 33 patients with histologic correlation, the gradation from cellular rim to relatively acellular center was gradual and often variable; therefore, no precise histologic correlation with signal intensity pattern on MR images was found (ie, no clear histologic boundary corresponded to the change in signal intensity seen on the long TRITE images). However, central low-signal-intensity areas within the tumor were found to represent areas of the tumor that were less cellular than the tumor rim. Any of these components may account, in part, for the hypointensity seen on MR images. Kovalikova et al (10) studied adenocarcinomas in animals and found that shortening of T2 was correlated with the appearance of coagulative necrosis within the himor. Similarly, Sillerud et al (11) found that central hypointensity and peripheral hyperintensity correlated precisely with coagulative necrosis and viable tumor rim, respectively, in an in vitro tumor model. Mature fibrosis in scars is usually hypointense, and predominantly fibrous tumors
332
#{149} Radiology
tend to be hypointense on long TRITE images. Mucin has a variable appearance on MR images and may be hypointense on long TRITE images if relatively dehydrated (12). It is not clear which of these elements may exert the greatest influence on the MR signal intensity of colorectal adenocarcinomas. We found no evidence to suggest that blood degradation products or calcification contributes significanfly to the central hypointense areas in the majority of cases. Given the frequency of desmoplastic reaction, necrosis, and mucin production of other common hepatic metastases (13), particularly those of gastrointestinal origin, and the range of MR patterns that we have found, it is unlikely that colorectal metastases will prove to have a specific MR imaging appearance. Fibrous pseudocapsules surrounding hepatic masses, manifested on MR images as areas of low signal intensity on short TRITE images and long TB/FE images, are a common feature of hepatocellular neoplasms and have been daimed to be a specific MR fealure for this diagnosis (2,8). In this study, low-intensity rims surrounding the metastatic tumor nodules were seen in 25% of lesions on long TRITE images. Seven lesions (4%) had rims evident on short TRIFE images, although these were usually thinner than those reported for hepatocellular neoplasms. Histologic correlation was available in 12 nodules and showed that these rims usually did not reflect the thick fibrous pseudocapsules reported for hepatocellular carcinoma. Rather, changes in the hepatic parenchyma around the tumors, predominantly compression of hepatic sinusoids and hepatocellular atrophy, variable amounts of fibrosis, as well as other histologic changes were present. Similar findings have been reported to be a frequent histologic feature in colon metastases to the liver (13). Rims around colorectal metastases due to hepatic compression have been demonstrated on enhanced CT scans as low-attenuation rings (14). Lin et al used microvascular injection to show that hepatic sinusoids adjacent to four of 15 metastases were compressed, forming an avascular zone in the liver around the tumor (14). We suggest that this compression may contribute to the lowsignal-intensity rims seen on MR images. In any case, these peripheral rims may not be distinguishable on MR images from those of hepatocellular carcinoma and adenoma. Thus,
this finding can not be assumed to be a specific MR feature of hepatocellular neoplasms. We conclude that metastases to the liver from colon or rectal adenocarcinomas have typical patterns of appearance on MR images. Metastatic tumor nodules were generally hyperintense on long TRITE images. Almost haif will show internal areas of low signal
intensity
within
the
hyperin-
tense nodule. The presence of these central changes within the tumor is size dependent and correlates to areas of desmoplastic stroma, cellular necrosis,
and
mucin
accumulation.
Pe-
ripheral halos of hyperintensity on long TRITE images were found to encompass the growing tumor margin and variable degrees of necrosis at the tumor rim, not peritumoral edema as described in the literature. Hypointense
peripheral
abnormalities specifically, renchyma,
rims
correspond
of hepatic compressed hepatocellular
fibrosis, and congested cent to the tumor edge.
to
parenchyma, hepatic paatrophy,
sinusoids
adja-
#{149}
References 1.
2.
ReimgJW, ams GW, tion with ogy 1989; Fthn PF,
Dwyer AJ, Miller Chang AE. Liver MR imaging at 0.5 170:149-151 Stark DD, Saini S,
ential diagnosis 3.
of ringed
D, FrankJA, Admetastases: detecand 1.5 T. Radiolet a!
The differ-
hepaticlesions
in
MR imaging. AJR 199t1; 154:287-290. Foley WD, KneelandJB, CatesJD, et a! Contrast optimization for the detection of focal hepatic lesions by MR imaging at 1.5 T. AJR 1987; 149:1155-1160.
4.
5.
6. 7.
8. 9.
Steinberg l-IV, Alarcon JJ, Bernardino ME. Focal hepatic lesions: comparative MR ima 1l at 0.5 and 1.5 T. Radiology 1990; 174:15
WittenbergJ, Stark DD, Forman BH, et at. Differentiation of hepatic metastases from hepatic hemangiomas and cysts by using MR imaging. AJR 1q88; 151:79-84. Stark DD, Bradley WG. Magnetic resonance imaging. St Louis: Mosby, 1988; 975. Itai Y, Ohtomo K, Kokubo T, Okada Y, Yamauclii T, Yoshida H. Segmental intensity differences in the liver on MR images: a sian of intrahepatic portal flow stoppage. Radiology 1988; l67:1-19. Rummeny E, Weissleder R, Stark DD, et at. Primary liver tumors: diagnosis by MR imaging. AJR 1989; 152:63-72. Muramatsu Y, Takayasu K, Morlyama N, et aL Peripheral low-density area of hepatic tumors:
CT-pathologic 10.
11.
12.
13.
14.
-
correlation.
Radiology
1986;
160:49-52. Kovalikova Z, Hoehn-Berlage MH, Gersonde K, Porschen R, Mittermayer C, FrankeR. Ae-dependent variation ofTi and T2 relaxation times of adenocarcinoma in mice. Radiology 1987; 164:543-548. Sillerud LO, FreyerJP, Neeman M, Mattingly MA. Proton NMR microscopy of multicellutar tumor spheroid morphology. Magn Reson Med 1990; 16:380-389. Dillon WP, Som PM Fullerton GD. Hypointense MR signal in chronically inspissated sinonasal secretions. Radiology 1990; 174:73Edinondson HA, Peters RL Neoplasms of the liver. In: Schiff L, Scuff ER, eds. Diseases of the liver. Philadelphia: Lippincott, 1982; 1101-1112. Lin C, Gustafson T, Hagerstrand I, Lunderquist A. CT demonstration of low density ring in liver metastasis. J Comput Assist Tomogr 1984; 8:450-452.
August1991
Outwater,
Hepatic of MR
MD
John
#{149}
E. Tomaszewski,
Colorectal Imaging and
Results of magnetic resonance (MR) imaging examinations for 76 patients with proved colorectal metastases to the liver were retrospectively reviewed. Signal intensity and morphologic patterns of 157 lesions were scored. The morphologic patterns were correlated with the histologic material obtained at segmental hepatic resections or excisional biopsies in 33 patients. In 154 lesions (98%) colorectal metastases to the liver were found to be hyperintense on MR images acquired with a long repetition time/echo time; in 77 lesions (49%) central low-signal-intensity changes were seen. Virtually all larger lesions demonstrated areas of low signal intensity within the himor. These areas were found to reflect histologic findings of coagulafive necrosis and desmoplasia within the tumor. Peripheral hyperintense halos around central hypointense areas encompassed the growing himor margin and variable degrees of cell necrosis. Hypointense peripheral rims, which correspond to abnormalities of surrounding hepatic parenchyma, were seen in 40 lesions (25%). This sign should not be assumed to represent the fibrous pseudocapsule of primary hepatic malignancy. No correlation between tumor grade (differentiation) and tumor morphology was observed. Index terms: Colon, neoplasms, 75.321 #{149}Liver neoplasms, MR studies, 76.1214 #{149}Liver neoplasms, secondary, 76.332 #{149}Rectum, neoplasms, 757.321
Radiology
‘From
1991; 180:327-332
the Departments
of Radiology
(E.O.,
H.Y.K.), Pathology (JET.), and Surgery (J.M.D.), Hospital of the University of Pennsylvania, 3400 Spruce St, Philadelphia, PA 19104. Received January 17,1991; revision requested February 12; revision received March 22; accepted April 1. Supported in part by NIH grant no. 1-ROlCA47354-0l. Address reprint requests to H.Y.K. ORSNA,
1991
MD
John
M. Daly,
#{149}
Metastases: Pathologic
MD
#{149} Herbert
importance of abdominal magnetic resonance (MR) imaging in the evaluation of liver metastatic disease is increasing, owing to its favorable sensitivity and specificity cornpared with those achieved with computed tomography (CT). To date, studies have focused predominantly on the efficacy of MR imaging in the detection of metastases, the optimal pulse sequences for their detection, and the differentiation of benign and malignant lesions based on morphologic patterns and signal intensity (15). As metastases in these studies tend to be treated in the aggregate, an understanding of the spectrum of MR appearances of metastases of any one histologic type is lacking. The variable MR imaging appearance of rnetastases to the liver was observed early in dinical studies of hepatic lesions. This variability may be seen even within one patient with one primary tumor (5). Although have
been
developed
the classification of MR pafterns hepatic mass lesions as an aid in tinguishing benign lesions from lignant (2,5), the microanatomic lates to the diverse metastases are not This retrospective
of dismacorre-
patterns of well characterized. MR study of he-
Patient
AND
mor and proved hepatic metastases. Five patients were exduded from analysis; three patients did not have images available for review, one patient had findings markedly obscured by recent surgical changes, and one patient with normal findings on MR images was subsequently shown to have metastasis at biopsy 3 months after the MR imaging examinalion. The remaining 76 patients comprised the study group. Medical records for each patient were reviewed for clinical data. The mean age of the patients was 62 years (range, 19-84 years);
46 were
men,
and
30 were
women.
The primary tumor was adenocarcinoma of the colon in 56 patients and rectal cardnoma in 20. Twenty-one patients had received systemic chemotherapy, usually fluorouracil, months to years prior to the MR imaging examination. Three of these patients received chemotherapy via an intraarterial infusion pump in the hepatic artery. Forty-nine patients had received no treatment for metastatic disease before the MR examination. In four cases, the previous treatment history could not be obtained.
for
patic metastases was performed with two goals in mind. First, we wished to establish the frequency and type of MR patterns in a single histologic type of metastasis, namely adenocarcinoma of the colon and rectum. Second, we have attempted to establish the microanatomic basis for these signal patterns.
PATIENTS
MD
Correlation Appearance’
T
schemata
Y. Kressel,
METHODS
Population
A computer search of records of patients who had undergone MR examinations on a 1.5-T system at this facility from January 1986 to October 1990 disdosed 81 patients with a history of a colorectal primary hi-
MR
Imaging
MR imaging system (Signa; waukee) with was an indusive tion extending pulse
sequences
was
performed
on a 1.5-T
GE Medical Systems, the body coil. Because retrospective
been
this
investiga-
back to 1986, a variety had
Mil-
of
employed.
Studies consisted of spin-echo images obtained in the axial plane with short repetition time (TR)/echo time (TE), long TR/ short TE, and long TRiTE sequences. Additional sequences were performed in many patients but were not used in the analysis of the studies. For the short TR[FE series, TRs ranged from 300 to 800 msec; 88% of the studies were acquired with TRs of either 400 or 600 msec. TEs varied from 25 to 12 msec. Images were obtained with two to four acquisitions and a 256 x 128 matrix, with a section thickness of 10 mm. The long TR images used a TR of 2,500
Abbreviations: tion time.
TE
=
echo time, TR
=
repeti-
327
msec
in 69 of the 76 patients;
nations
seven
exami-
were
performed with longer TRs. Symmetric TEs of 40 and 80 msec were used in 68 of the 76 patients; asymmetric echoes of 20-30 and 80 msec were used in the remaining patients. The section thickness was 5-10 mm, and an acquisition trix of 256 x 128, with two acquisitions,
ma-
area) for correlation of the pathologic and MR image features. Because this was a retrospective study, precise localization of the tissue block to the imaged tumor was not possible; however, in each case the tumor edge/liver interface could be exammed. Each of these histologic specimens was retrospectively reviewed by one pa-
was used in most cases. Respiratory cornpensation by means of reordering of phase-encoding steps and gradient-mornent nulling was generally employed. In eight patients, hepatic resection specimens were imaged with MR prior to fixation. A 10-16-cm field of view, 5-mm section thickness, and spin-echo sequences of 400-600/20 and 2,500/40, 80 were used to image the specimens. One image per patient was retrospeclively analyzed by two of the authors. If
O.T.) for specific features in the and surrounding liver. These induded the relative amounts of desmoplastic reaction, tumor necrosis, mucin accumulation within the tumor, and tumor
patients
In the 76 patients, 147 individual tumor nodules were identified on the MR images, and 10 patients had metastases that were too numerous to count or were confluent. The latter
had
more
than
one
image
avail-
able, the study performed dosest in time to the liver biopsy was chosen for analysis. Every lesion in the liver was examined for size, morphology, and signal intensity. Lesions that became progressively more intense
relative
to liver
on
heavily
segmental when
hepatic
lesion
or area
of tumor.
such as adenopathy edema were noted
or
present.
The size of each tumor mass was measured at its greatest diameter in the axial plane, with the accompanying perpendicular diameter. The size is expressed as the product of these diameters in square centimeters.
Pathologic
Review
The presence
of metastatic
disease
these patients was established biopsy in 67 patients, needle cytology in six patients, and
the hepatic
metastases
biopsy of the hepatic patients. Thirty-three
tumor
nodules)
tectomy,
wedge
resection,
of
aspiration palpation
metastases patients
underwent
in
by means
at surgery
of
without in three (with 36
partial
hepa-
or excisional
biopsy of their metastasis to yield enough tissue (approximately 1-4-cm2 section 328
Radiology
#{149}
tumor
grade.
RESULTS MR Imaging Metastases
lesions
T2-
weighted sequences (2,500/120-160) and met standard morphologic criteria for cysts or hemangiomas were exduded from analysis. Signal intensity of the components of the individual lesions was subjectively scored on a seven-point scale (1 = very high signal intensity, equal to that of cerebrospinal flu. on a long TRiTE sequence or fat on a TI-weighted sequence; 2 = hypetinten compared with liver; 3 = slightly hyperint.nse compared with liver; 4 = isointense with liver or not seen; 5 = slightly hypointense compared with liver; 6 = hypointense compared with liver; 7 = very low signal intensity, equal to that of simple fluid on a short TRIFE sequence). Lesions that were confluent or too numerous to count were treated for the purposes of scoring as a solitary lesion having the signal characteristics of the largest Ancillary findings
thologist
were
Characteristics
treated
of
as solitary
or not
seen,
lesions
as opposed
to
three of 157 (2%) on the long TR/TE images (Table 1). Part of the failure to visualize these lesions on the short TRIFE images, however, may be due to the thicker sections used in some cases for this sequence, as well as the TEs of up to 25 msec. Intermediateweighted sequences (short TFllong TR images) rendered more lesions isointense or slightly hyperintense to liver. Hence, this sequence appeared to be inferior to that of long TRTFE images for the detection of metastatic deposits.
Of the 157 lesions, strated central areas lower signal intensity higher-intensity
tumor
77 (49%) that were relative edge
demonof to the on
inhomogeneity was found patients with a history of syschemotherapy (44% preva-
temic lence). In contrast, focal areas of very high signal intensity on the T2-weighted images correlated with areas of very low signal intensity on the TIweighted images. These signal characteristics are indicative of fluid, and liquefactive necrosis appeared to account for the high signal intensity (Fig 1). This central high intensity occurred in 24 of 157 lesions (15%) and was seen only in larger lesions (Fig 2). In 21 of 24 cases (88%), areas of very high intensity within the tumor were surrounded
for the purposes of analysis, yielding a total of 157. As expected, the predominant signal intensity of the himors was moderately low relative to normal liver on short TRITE sequences and moderately high on long TR/TE images. On short TRiTE images, 25 of 157 lesions (16%) were isointense
intensity in those
the
T2-weighted images (Fig 1). These areas varied from small foci to confluent central areas of low signal intensity surrounded by a rim of higher signal intensity. Large lesions were more likely to demonstrate inhomogeneous signal intensity than smaller lesions (Fig 2), although even some small lesions showed centers of low signal intensity. No correlate to these areas on the TI-weighted images could be found, that is, both the lowintensity center and the high-intensity rim showed uniformly low intensity on the short TRIFE images. No higher frequency of central signal-
by
areas
of low
intensity.
Only four lesions contained foci of high intensity on short TRIFE images. These areas were also of high intensity on the T2-weighted images, consistent with intratumoral hemorrhage or highly proteinaceous fluid. In 40 lesions (25%), a thin rim of homogeneous low signal intensity surrounded the high-signal-intensity tumor on long TR/short TE and long TRTFE images (Fig 3). The rim varied from 2 to 8 mm thick and increased the contrast and distinctness of the border between the tumor and the surrounding liver. In seven lesions, a low-signal-intensity rim appeared on the Ti-weighted observed only
isointense
images. in lesions
or nearly
This was showing
signal images in contrast to lesions with a low-signal-intensity rim (Figs 4, 5). Similar to lesions with the low-signal-intensity characteristics
centers,
on
these
isointense
short
TR/TE
low-intensity
rims
pre-
dominated also seen sions. slightly
in larger lesions but were around some smaller leIn two patients a faint rim of high
intensity
surrounded
the
low-intensity tumor on short TR/FE images. Long TR/TE images in five patients demonstrated discrete wedge-shaped areas of slightly high signal intensity corresponding to vascular territories. These areas were associated with centrally located tumors in a position to compress or ocdude portal vessels or segmental bile ducts.
Histologic Examination
Findings of the
histologic
sec-
tions from 33 patients demonstrated a generally stereotypic histologic pattern. Viable tumor cells tended to cluster within several millimeters of the periphery of the nodule and were admixed with small areas of mucin and cell necrosis. More centrally, large August1991
Table 1 Predominant Metastases
Signal Intensity of 157 Related to Pulse Sequence Pulse Sequence
Signal Intensity Rating
5,
,
r”v: ,
L
.p
0
Short TR/TE
Long TR/ Short TE
1
0
0
7
2 3 4 5 6 7
0 0 25 43 88 1
53 70 32 1 1 0
119 27 3 0 1 0
100 95 90 85 80 75 70
.c
w .
.2
ll 0
C, C-
65 60 55 50 45 40 35 30 25 20 15 10
I
H1
NR 3
5
Lesion
Figure d.
C.
Figure
1. Lowand high-internal-signal-intensity changes. (a) Long TRITE (2,500/80) shows hyperintense edge of tumor, confluent area of low signal intensity internally, punctate region of high signal intensity centrally. (b) Axial (2,500/80) image obtained surface coil of left lobectomy specimen (c) shows the well-demarcated hypenntense rim corresponding to the histologic finding of advancing cellular tumor rim (arrows
(c) Large-mount
section
of right half of tumor
(trichrome
stain).
Areas
of low signal
image and with a tumor
in c). intensity
2.
Signal
10
30
Size
intensity
100
I 200
300
TNTC
(cm
patterns
related
to
lesion size. Lesion size is expressed as the product of maximum perpendicular diameters in the axial plane. Dotted bars represent lesions with foci of central low signal intensity. Patterned bars represent lesions with low-signal-intensity rims. Black bars repre-
sent lesions sity. TNTC
with =
central
too
high
numerous
signal
inten-
to count.
on MR stained represent
stain
images correspond to the areas of coagulative necrosis (arrowheads), which are dark red, and fibrosis, which are stained blue. High-signal-intensity foci in a and b accumulations of cell debris (*) in the center of the tumor. (d) Use of prussian blue fails to demonstrate any hemosiderin deposits to account for the low-intensity areas. L = liver.
Long TR/TE
areas
of coagulative
plasia, and, accumulation Viable
tumor
mediate were the
cells
edge usually
All tumor and
to represent
desmo-
from
tumor
sparse
compared and
nodules
the
without
low-intensity
long
TRiTE
nent from
that the
and
images.
was tumor
centers
The
tissue
on
compo-
most prevalent away edge was determined
correlated
with
the
MR
imaging
signal pattern. These results are summarized in Table 2. Although other central tissue components were usually present, focal and confluent centers of low intensity within the tumor on the long TR/TE images appeared Volume
180
Number
#{149}
2
cellular
from no
histologic feature absolutely discriminated between tumor nodules with and
more rim.
areas
At histologic
some
Therefore,
of more
marked
near
the
tumor
examinations,
low-
signal-intensity rims around nodules on the T2-weighted
debris.
desmoplasia,
production.
areas
desmoplastic reaction and coagulative necrosis. These features blended with
with cell
displayed
necrosis,
im-
nodule
of fibrosis
of cell mucin
away
of the
amount
degree
necrosis,
to varying degrees, mucin predominated (Fig 1).
12 of 33 patients
represent hepatic
tumor images
were
pathologic parenchyma
found
changes adjacent
to
in the to the
tumor.
Compression
of hepatic
paren-
chyma atrophy,
and sinusoids, and fibrosis
hepatocellular occurred most
frequently (Fig 3), with foci of hemorrhage, inflammation, and congested sinusoids
sometimes
demonstrates peritumoral
encountered. ity
as determined
specimens-that from normal
one fibrotic
The
seen.
of the
Figure
more
reactions
3d
marked that
zone
of abnormal-
from
the
histologic
is, the transition liver to tumor edge-
ranged
from
0.3 to 8.0 mm
wider in lesions signal-intensity (mean, 2.7 mm) nodules without images There
(mean, was no
and
was
demonstrating a lowrim on MR images than in those tumor visible rims on MR 0.7 mm; P < .005). histologic
evidence
of
extracellular fluid accumulations around the tumor to indicate the presence of vasogenic edema that could contribute to the signal intensity pattern as seen on MR images. No
histologic
the
faint
correlate
was
found
high-signal-intensity
on Ti-weighted seen around
to
rims
images, which were metastases. Iron of tumors from five patients three
staining was performed and dence of significant in the tumors. were observed mens.
we
Enlarged fled the
in nine presence
revealed no eviiron deposition
No calcium on histologic
lymph
nodes
deposits speci-
were
identi-
patients. In five of these, of lymphatic metastasis
was confirmed at biopsy. tients, lymphadenopathy
In two pawith lipoRadiology
329
#{149}
Figure
3.
Low-signal-intensity
rim
on
long
TR/I’E images. (a) Long TRITE (2,500/80) image demonstrates a metastasis in the right lobe of the liver with a 2-3-mm rim of low signal intensity (arrows). (b) The rim of compressed hepatic parenchyma (between arrowheads) adjacent to the tumor edge corresponds to the low-intensity rim seen in a.
(Hematoxylin-eosin
stain;
original
magnifica-
tion, x2.) (c) A lesion in another patient shows a similar low-intensity rim on this spin-echo (2,500/80) image, representing the histologic finding (d) of a rim of fibrous tissue (arrowheads) between hepatic parenchyma (L) and tumor (fl. (Hematoxylin-eosin stain; original magnification, x2.)
granulomas without found at biopsy. The
metastasis other two
was pa-
tients did not have surgical confirmation. Two patients had periportal lymph node metastases that were not seen
on
MR
moderately
slides diswere in eight
to poorly
differ-
entiated in 21 patients, and poorly differentiated in four. No definite terns on MR images corresponded these
b.
images.
Review of the histologic closed that the carcinomas moderately differentiated patients,
a.
histologic
grades,
as signal
patto inho-
mogeneity was similar for these groups. Similarly, no definite correlation emerged of treatment
between a prior with chemotherapy
the histologic aging pattern. bid high
features, signal
history and
findings or the MR imCarcinomas with colhowever, intensity
on
images
in three
of four
paucity
of focal
low-intensity
showed very T2-weighted
cases,
with
a
areas ‘
(Fig 5).
DISCUSSION ‘.34’
Colon and rectal adenocarcinomas are among the most frequent primary tumors to metastasize to the liver and are the most amenable to surgical resection. For MR imaging to play an important
role
in the
management
these cases it must demonstrate number of metastases present
of
the with
t..
high sensitivity, enable distinction between benign and malignant lesions with high specificity, and help establish the extent of each lesion to
determine ing
its resectability.
at 1.5 T, the
long
TRITE
For imagimage
has
the highest sensitivity for hepatic metastases (3) and provides the critical morphologic and signal intensity information to help distinguish benign from malignant lesions (5). With these sequences, however, colorectal metastases manifest logic and signal
Accordingly, 330
#{149} Radiology
a variety intensity
a knowledge
of morphopatterns.
of the
4
b.
a.
Figure
4. Low-signal-intensity shows isointense tumor mass fibrosis and hepatic compression eosin stain).
rim
on short TRITE with thin hypointense adjacent to tumor
range of MR imaging patterns and their meaning with long TRIfE sequences is important. Review of the MR findings in the 157
metastases
in this
series
supports
the findings of others that long TRITE images are superior to short TRITE or
image. (a) Ti-weighted (400/15) image rim (arrows). (b) Section demonstrates edge (between arrows) (hematoxylin-
long
TR’ short
TE images
in demon-
strating metastases at 1.5 T (3). Approximately one-quarter and one-
third, respectively, of the individual lesions were isointense on the latter sequences. Only one-half, however, were relatively homogeneous in inAugust
1991
Table 2 Correlation of Histologic Findings Long TRITE MR Imaging Findings Tumor Nodules from 33 Patients Internal
and in
Signal
Intensity Tissue
Component
Coagulative (n=16)
High
Low
necrosis
Fibrosis (i = 9) Tumor cefls
(n=4) Mucin/cell (n=3)
7
9
0
9
4
0
2
1
1
0
debris
Liquefactive
necrosis
(n=1)
with peritumoral edema. We found no examples of low-signal-intensity -
I
-
.
nodules
without
Second,
the
a hyperintense
size
of the
rim.
metastases
found at surgery corresponded size of the hyperintense mass weighted
MR
pointense was often
central area alone, very small relative
metastases
on
sharply tumor
from
5.
Colloid carcinoma. (a) Short TRITE (400/20) image shows a thin hypointense rim surrounding an isointense tumor with a hypointense center. (b) Histologic specthe edge of the tumor (T) shows a rim of compressed hepatic parenchyma associ-
(arrowheads) imen
from
ated with obliteration
of mild steatosis
present
stain). (c) Long TRITE tensity corresponding histologic section (d).
(2,500/80) image to accumulations
of the same
tensity
on
degree
of relative
sity
was
many
long
of lower
TR/TE
images.
central
areas foci
intensity
a thin areas
areas
were
lower-intensity
of signal change were correlated
metastasis;
to suggest these patwith hypoin-
virtually
small,
cen-
surrounded rim. The intensity
surrounded areas.
region
surrounded
from
to large
hyperintense of high signal
usually types mor
ranging
of nodules;
a confluent
by the hyperintense rim the “halo” sign. However, terns form a continuum, tral
by central (fluid)
by these Both
of these
180
demonstrates
the marked
(*) and cell debris within
hyperin-
the tumor
on the
#{149} Number
if the
images. The clinical significance this is that this peripheral zone
growth.
It has been suggested that sign on T2-weighted images sents ule
a low-intensity surrounded
the halo repre-
metastatic by
edema
nod-
within
the
liver, based on clinical studies of metastases of various primaries (5,6). We found no evidence to suggest that this mechanism is responsible for the signal intensity patterns of colorectal metastases.
This
conclusion
is based
metastases a low-intensity
sity rind, of nodules
that
includes center
lesions and
with
high-inten-
rather than a distinct group wholly of low intensity
a
area
We conclude, therefore, majority of cases, concentric edema does not contribute
to tumor
all of the
2
liver,
to occur
significantly
on our findings that, first, there is a spectrum of appearances of colorectal
largest
surrounding
that these internal changes in signal intensity represent a spectrum of internal histologic changes related
within the tuto the size of the
tumors showed some degree of relative high- or low-signal-intensity changes within the tumor. We sugVolume
in the liver (L) (hematoxylin-eosin
lesion
gest
Some
hypointen-
by 49%
had
signal
of mucin
elsewhere
images
hyperintense of
hyperintensity actually represents hepatic parenchymal edema. Lastly, the depth of histologic abnormality in the surrounding hepatic parenchyma around the nodules that were exammed histologically was insufficient to account for the peripheral hyperintensity seen in most of those nodules demonstrating a low-signal-intensity center. in the hepatic
displayed
of these
tense
..-
the
hy-
which to the
region of hylow-signalaround some
the
unlikely
to the T2-
to the
T2-weighted
demarcated
finding
Figure
not
size of the surrounding perintensity. Third, the intensity rims observed
*
/
images,
on
signal tastases
to the
intensity on long
perintensity represent
ring
that,
patterns
of
of colorectal meTRITE spin-echo
should be assumed tumor for the purposes
surgical
planning
and
tumor
MR of of hy-
to of
volumet-
nc analysis. This
is in contrast
shaped
intensity generally the tumor
areas
to the
of slightly
on long extended toward
TRITE
wedge-
high
signal
images
peripherally the hepatic
that from capsule.
Itai et al (7) noted the former finding in six of 73 cases of primary and secondary malignant hepatic tumors and showed struction Rummeny
that intrahepatic portal obmay account for the finding. et al (8) suggested that
Radiology
#{149} 331
lymphatic obstruction may play a role and that the finding was not associated with metastases. We observed this pattern, however, in five patients. Thus, although uncommon, it can occur in association with colorectal metastases. Muramatsu et al (9) have classified colorectal metastases in the liver into three histologic patterns: those demonstrating a completely peripheral distribution of tumor cells and a fibrotic center; those with a predominantly peripheral distribution of himor cells with mixed desmoplasia and carcinoma in the center; and those with mixed carcinoma cells, coagulative necrosis, and fibrosis throughout the mass. Although we found these distinctions somewhat arbitrary, as virtually all of the histologic specimens we reviewed had a predominantly peripheral distribution of carcinoma and very few completely lacked carcinoma cells centrally, this descriptive classification does serve to generally characterize the internal structure of these lesions accurately. Internal desmoplastic reaction is often a dominant feature of these lesions and may contribute to the central areas of low signal intensity seen in some tumors. We found that these areas usually contain a considerable amount of necrotic cellular debris interspersed with the fibrosis, although they have been shown to be perfused on late-enhanced CT scans (9). In the nodules from the 33 patients with histologic correlation, the gradation from cellular rim to relatively acellular center was gradual and often variable; therefore, no precise histologic correlation with signal intensity pattern on MR images was found (ie, no clear histologic boundary corresponded to the change in signal intensity seen on the long TRITE images). However, central low-signal-intensity areas within the tumor were found to represent areas of the tumor that were less cellular than the tumor rim. Any of these components may account, in part, for the hypointensity seen on MR images. Kovalikova et al (10) studied adenocarcinomas in animals and found that shortening of T2 was correlated with the appearance of coagulative necrosis within the himor. Similarly, Sillerud et al (11) found that central hypointensity and peripheral hyperintensity correlated precisely with coagulative necrosis and viable tumor rim, respectively, in an in vitro tumor model. Mature fibrosis in scars is usually hypointense, and predominantly fibrous tumors
332
#{149} Radiology
tend to be hypointense on long TRITE images. Mucin has a variable appearance on MR images and may be hypointense on long TRITE images if relatively dehydrated (12). It is not clear which of these elements may exert the greatest influence on the MR signal intensity of colorectal adenocarcinomas. We found no evidence to suggest that blood degradation products or calcification contributes significanfly to the central hypointense areas in the majority of cases. Given the frequency of desmoplastic reaction, necrosis, and mucin production of other common hepatic metastases (13), particularly those of gastrointestinal origin, and the range of MR patterns that we have found, it is unlikely that colorectal metastases will prove to have a specific MR imaging appearance. Fibrous pseudocapsules surrounding hepatic masses, manifested on MR images as areas of low signal intensity on short TRITE images and long TB/FE images, are a common feature of hepatocellular neoplasms and have been daimed to be a specific MR fealure for this diagnosis (2,8). In this study, low-intensity rims surrounding the metastatic tumor nodules were seen in 25% of lesions on long TRITE images. Seven lesions (4%) had rims evident on short TRIFE images, although these were usually thinner than those reported for hepatocellular neoplasms. Histologic correlation was available in 12 nodules and showed that these rims usually did not reflect the thick fibrous pseudocapsules reported for hepatocellular carcinoma. Rather, changes in the hepatic parenchyma around the tumors, predominantly compression of hepatic sinusoids and hepatocellular atrophy, variable amounts of fibrosis, as well as other histologic changes were present. Similar findings have been reported to be a frequent histologic feature in colon metastases to the liver (13). Rims around colorectal metastases due to hepatic compression have been demonstrated on enhanced CT scans as low-attenuation rings (14). Lin et al used microvascular injection to show that hepatic sinusoids adjacent to four of 15 metastases were compressed, forming an avascular zone in the liver around the tumor (14). We suggest that this compression may contribute to the lowsignal-intensity rims seen on MR images. In any case, these peripheral rims may not be distinguishable on MR images from those of hepatocellular carcinoma and adenoma. Thus,
this finding can not be assumed to be a specific MR feature of hepatocellular neoplasms. We conclude that metastases to the liver from colon or rectal adenocarcinomas have typical patterns of appearance on MR images. Metastatic tumor nodules were generally hyperintense on long TRITE images. Almost haif will show internal areas of low signal
intensity
within
the
hyperin-
tense nodule. The presence of these central changes within the tumor is size dependent and correlates to areas of desmoplastic stroma, cellular necrosis,
and
mucin
accumulation.
Pe-
ripheral halos of hyperintensity on long TRITE images were found to encompass the growing tumor margin and variable degrees of necrosis at the tumor rim, not peritumoral edema as described in the literature. Hypointense
peripheral
abnormalities specifically, renchyma,
rims
correspond
of hepatic compressed hepatocellular
fibrosis, and congested cent to the tumor edge.
to
parenchyma, hepatic paatrophy,
sinusoids
adja-
#{149}
References 1.
2.
ReimgJW, ams GW, tion with ogy 1989; Fthn PF,
Dwyer AJ, Miller Chang AE. Liver MR imaging at 0.5 170:149-151 Stark DD, Saini S,
ential diagnosis 3.
of ringed
D, FrankJA, Admetastases: detecand 1.5 T. Radiolet a!
The differ-
hepaticlesions
in
MR imaging. AJR 199t1; 154:287-290. Foley WD, KneelandJB, CatesJD, et a! Contrast optimization for the detection of focal hepatic lesions by MR imaging at 1.5 T. AJR 1987; 149:1155-1160.
4.
5.
6. 7.
8. 9.
Steinberg l-IV, Alarcon JJ, Bernardino ME. Focal hepatic lesions: comparative MR ima 1l at 0.5 and 1.5 T. Radiology 1990; 174:15
WittenbergJ, Stark DD, Forman BH, et at. Differentiation of hepatic metastases from hepatic hemangiomas and cysts by using MR imaging. AJR 1q88; 151:79-84. Stark DD, Bradley WG. Magnetic resonance imaging. St Louis: Mosby, 1988; 975. Itai Y, Ohtomo K, Kokubo T, Okada Y, Yamauclii T, Yoshida H. Segmental intensity differences in the liver on MR images: a sian of intrahepatic portal flow stoppage. Radiology 1988; l67:1-19. Rummeny E, Weissleder R, Stark DD, et at. Primary liver tumors: diagnosis by MR imaging. AJR 1989; 152:63-72. Muramatsu Y, Takayasu K, Morlyama N, et aL Peripheral low-density area of hepatic tumors:
CT-pathologic 10.
11.
12.
13.
14.
-
correlation.
Radiology
1986;
160:49-52. Kovalikova Z, Hoehn-Berlage MH, Gersonde K, Porschen R, Mittermayer C, FrankeR. Ae-dependent variation ofTi and T2 relaxation times of adenocarcinoma in mice. Radiology 1987; 164:543-548. Sillerud LO, FreyerJP, Neeman M, Mattingly MA. Proton NMR microscopy of multicellutar tumor spheroid morphology. Magn Reson Med 1990; 16:380-389. Dillon WP, Som PM Fullerton GD. Hypointense MR signal in chronically inspissated sinonasal secretions. Radiology 1990; 174:73Edinondson HA, Peters RL Neoplasms of the liver. In: Schiff L, Scuff ER, eds. Diseases of the liver. Philadelphia: Lippincott, 1982; 1101-1112. Lin C, Gustafson T, Hagerstrand I, Lunderquist A. CT demonstration of low density ring in liver metastasis. J Comput Assist Tomogr 1984; 8:450-452.
August1991
