Takashi Koyama, MD
Joel G. Fletcher, MD
C. Daniel Johnson, MD
Mark S. Kuo, MD
Kenji Notohara, MD
Lawrence J. Burgart, MD
Index terms:
Liver neoplasms, 761.322
Liver, nodules, 761.322
Sarcoma, 761.322
Published online before print
10.1148/radiol.2223010877
Radiology 2002; 222:667– 673
1
From the Department of Diagnostic
Imaging and Nuclear Medicine, School
of Medicine, Kyoto University, Japan
(T.K.); Departments of Radiology (J.G.F.,
C.D.J.) and Anatomic Pathology (K.N.,
L.J.B.), Mayo Clinic, 200 First St SW,
Mayo East 2B, Rochester, MN 55905;
and Department of Radiology, Scotts-
dale Medical Imaging, Ariz (M.S.K.).
Received May 2, 2001; revision re-
quested June 8; revision received and
accepted September 7. Address cor-
respondence to J.G.F. (e-mail: fletcher
.joel@mayo.edu).
© RSNA, 2002
Author contributions:
Guarantors of integrity of entire study,
T.K., J.G.F.; study concepts and de-
sign, T.K., J.G.F.; literature research,
T.K.; clinical studies, T.K., J.G.F.; ex-
perimental studies, all authors; data
acquisition and analysis/interpretation,
T.K., J.G.F., C.D.J.; manuscript prepara-
tion and definition of intellectual con-
tent, all authors; manuscript editing,
T.K., J.G.F., C.D.J.; manuscript revision/
review, all authors; manuscript final ver-
sion approval, T.K., J.G.F., C.D.J.
Primary Hepatic
Angiosarcoma: Findings at CT
and MR Imaging1
PURPOSE: To evaluate and describe cross-sectional imaging findings in patients
with pathologically confirmed primary hepatic angiosarcoma.
MATERIALS AND METHODS: Findings from imaging examinations in 13 patients
with pathologically confirmed primary hepatic angiosarcoma were retrospectively
reviewed (computed tomographic [CT] images obtained in 10 patients and mag-
netic resonance [MR] images obtained in five patients were available for review).
Two gastrointestinal radiologists evaluated lesion number, size, attenuation and
signal intensity characteristics, and the pattern and degree of contrast material
enhancement. Medical records were reviewed for clinical features associated with
angiosarcoma.
RESULTS: Angiosarcoma appeared as multiple nodules (n ⫽ 6), as dominant masses
(n ⫽ 6), or as a diffusely infiltrating lesion (n ⫽ 1). Multiple nodules were hypoat-
tenuating at unenhanced and contrast material– enhanced CT (six of six patients).
When dominant masses were encountered at MR imaging, T2-weighted MR imag-
ing demonstrated heterogeneous internal architecture (four of four patients) similar
to that of hepatocellular carcinoma. Multiphase contrast-enhanced CT and MR
images showed dominant masses to have heterogeneous and progressive enhance-
ment (three of three patients). Clinical features associated with angiosarcoma
included splenic metastases (six of 13 patients), thrombocytopenia (seven of 13
patients), disseminated intravascular coagulation (four of 13 patients), and hemo-
lytic anemia (three of 13 patients).
CONCLUSION: Primary hepatic angiosarcoma exhibits a spectrum of appearances
that reflect its varied pathologic features.
© RSNA, 2002
While primary hepatic angiosarcoma accounts for only 2% of primary hepatic tumors, it
is the most common malignant mesenchymal tumor of the liver (1– 4). Hepatic angiosar-
coma portends a poor prognosis, and most patients die within a year of diagnosis (3).
When the lesion is confined to one lobe of the liver without any metastatic lesions,
however, it is resectable. Because of the vascularity of the lesion, percutaneous biopsy has
been reported as treacherous, complicated by massive hemorrhage in a minority of
patients (3).
Early reports of hepatic angiosarcoma focused on its association with environmental
carcinogens, such as thorium dioxide (Thorotrast), arsenic, and vinyl chloride, but expo-
sure to these agents is now rare. Most of these tumors occur either in the absence of known
risk factors or with cirrhosis (2).
Various appearances of hepatic angiosarcoma on computed tomographic (CT) images
have been described in case reports and in a few small series (5–11). On CT images,
angiosarcoma has been known to simulate benign hemangioma or metastases (7,11). The
appearance of this tumor at magnetic resonance (MR) imaging is described in only a few
sporadic case reports in the English-language literature, so its MR features have not been
clearly delineated (11–13). The purpose of our study was to evaluate and describe cross-
sectional imaging findings in patients with pathologically confirmed primary hepatic
angiosarcoma.
667
MATERIALS AND METHODS
After receiving approval from our institu-
tional review board, we reviewed cross-
referenced records from May 1984 to Jan-
uary 2001 in the departments of radiology
and pathology at the Mayo Clinic and
identified 13 patients with pathologically
confirmed angiosarcoma. Patients were
not individually asked for consent to be
included in this study, but each patient
in the study did agree to the retrospective
use of medical records and images for
research purposes during treatment at
our institution. These 13 patients, in-
cluding eight men and five women, con-
stituted our study population. Patient age
ranged from 37 to 84 years (mean age,
66.7 years ⫾ 10.9 [SD]). CT images were
available in 10 patients, and MR images
were available in five patients. Two pa-
tients had both CT and MR images avail-
able for review.
All patients had biopsy-proven hepatic
angiosarcoma (nine patients underwent
CT-guided biopsy, two underwent surgi-
cal resection, and four underwent explor-
atory laparotomy). CT-guided biopsy was
unsuccessful in two patients (requiring
subsequent exploratory laparotomy), and
substantial bleeding was not observed in
the immediate postprocedure period in
any of the patients who underwent percu-
taneous biopsy.
Unenhanced CT images were obtained
in six patients, and contrast material– en-
hanced CT images were obtained in eight
patients with the use of intravenous iodin-
ated contrast material. Six patients re-
ceived 140–150 mL of iopamidol (Isovue
300; Bracco Diagnostics, Princeton, NJ),
and two patients received 150 mL of iothal-
amate meglumine (Conray 60; Mallinck-
rodt Medical, St Louis, Mo). Regarding the
contrast-enhanced examinations, one pa-
tient underwent imaging in the arterial
phase, seven underwent imaging in the
portal venous phase, and two underwent
delayed imaging. Section thickness ranged
between 5 and 10 mm.
Transverse T1- and T2-weighted MR
images were obtained in five patients. A
conventional spin-echo technique was
used to obtain T1-weighted images (rep-
etition time msec/echo time msec, 250 –
660/14 –20). Regarding T2-weighted im-
ages, a fast spin-echo technique (4,000 –
8,571/105–140) was used in two patients,
and a conventional spin-echo technique
(2,000 –2,500/100 –120) was used in three
patients. Dynamic T1-weighted three-di-
mensional fast spoiled gradient-echo im-
ages were obtained in two patients after
668 䡠 Radiology 䡠 March 2002
intravenous administration of 10 mL of
gadopentetate dimeglumine (Magnevist;
Berlex Laboratories, Wayne, NJ) or 16 mL
of gadoteridol (ProHance; Bracco Diag-
nostics), respectively.
Two gastrointestinal radiologists (T.K.,
J.G.F., C.D.J.) examined lesion size (di-
ameter in centimeters), number, and lo-
cation, as well as characteristics such as
hemorrhage, necrosis, and sites of metas-
tasis, and formed a consensus interpreta-
tion. Tumor attenuation (for the entire
tumor and for the foci of enhancement
within the tumor) was visually graded
as greater than, less than, or equal to
the attenuation of the surrounding he-
patic parenchyma and the aorta. En-
hancement patterns within tumors were
characterized as focal regions of enhance-
ment, irregularly shaped regions of en-
hancement, or rim enhancement. On MR
images, the signal intensity characteris-
tics of the lesion were compared with
those of the surrounding liver paren-
chyma and the muscles, and the pattern
of enhancement was also characterized,
as described above. Pathologic findings
in surgical specimens were compared
with imaging findings (T.K., K.N., L.J.B.).
Point-to-point comparison was not pos-
sible, however, since only two patients
underwent complete surgical resection.
Clinical records of patients were re-
viewed (T.K., J.G.F.) regarding clinical
presentation, preexisting diseases, ab-
normal findings of laboratory tests, and
history of exposure to environmental
carcinogens. Clinical histories were re-
viewed for history of exposure to Thoro-
trast, vinyl chloride, anabolic steroids, or
arsenic, and for history of hemochroma-
tosis, chronic hematoma, or pyothorax.
The following values were obtained from
laboratory tests and recorded at diagnosis
of angiosarcoma (n ⫽ 12) or at presenta-
tion to our institution (n ⫽ 1), when ap-
plicable: hemoglobin (n ⫽ 13), hemato-
crit (n ⫽ 11), platelets (n ⫽ 13), D-dimer
(n ⫽ 4), and fibrinogen (n ⫽ 4).
Clinical records of patients were also
reviewed for sites of metastasis. All 13
patients underwent either chest CT (n ⫽
9) or chest radiography (n ⫽ 13) to doc-
ument pulmonary metastases. Splenic
metastases were assessed either with sur-
gical specimens (n ⫽ 2) or with CT or MR
imaging of the abdomen (n ⫽ 11), as
previously described. Bone metastases
were noted when present on any image
(CT, MR, or plain radiographic). All pa-
tients underwent MR imaging or CT of
the abdomen, nine underwent CT of the
chest, seven underwent CT of the pelvis,
and one underwent a radiographic, met-
astatic bone survey.
RESULTS
Clinical Information
Patients most frequently presented
with upper-quadrant pain (n ⫽ 7), ab-
dominal discomfort (n ⫽ 7), anorexia
(n ⫽ 4), or weight loss (n ⫽ 3). Findings of
laboratory tests revealed that seven pa-
tients had thrombocytopenia, four of
whom had disseminated intravascular
coagulation. All four patients with dis-
seminated intravascular coagulation had
elevated D-dimer and low fibrinogen lev-
els. Eight patients had anemia, including
three with microangiopathic hemolytic
anemia and one with known aplastic
anemia. Several patients had medical his-
tories related to their development of an-
giosarcoma. Two patients had a history
of exposure to Thorotrast. One patient
had a chronic organized hematoma along
the lateral surface of the liver (a compli-
cation of percutaneous transhepatic bili-
ary drainage 5 years prior to presenta-
tion), with a massive angiosarcoma in
the vicinity of the hematoma. Another
patient had abused anabolic steroids and
acquired transfusion-induced hemochro-
matosis subsequent to his known aplastic
anemia. Another patient had a husband
who worked with arsenic and lead-based
pesticides. None of the patients in our
study had a history of exposure to vinyl
chloride.
Imaging Findings
The number and size of hepatic tumors
were based on CT and MR imaging find-
ings. Six patients had multiple small nod-
ules, usually measuring less than 3 cm in
diameter, scattered within both lobes of
the liver. Five patients had a large domi-
nant mass measuring 8 –14 cm (three in
the right lobe, one in the left lobe, and
one in both lobes). Four of these five
patients had other intrahepatic lesions
that measured less than 3 cm. One pa-
tient had two large focal lesions (one in
each lobe) measuring 7–9 cm, and one
patient had a diffusely infiltrating tu-
mor throughout the liver.
Seven patients (54%) had metastatic
lesions. Six (46%) had splenic metasta-
ses, while three (23%) had lung metas-
tases. Three (23%) also had bone metasta-
ses, but only eight patients underwent
either a metastatic bone survey or CT of
the entire chest, abdomen, and pelvis. On
unenhanced CT images, all lesions were
hypoattenuating compared with normal
Koyama et al

Figure 1. Images show angiosarcoma in a 62-
year-old man. (a) Transverse contrast-enhanced
CT scan shows multiple hypoattenuating liver
lesions, some with foci of enhancement (arrow-
heads), which are of decreased attenuation com-
pared with the aorta. (b) Photomicrograph
shows dilated, cavernous vascular channels (ar-
rows) lined with neoplastic cells. (Hematoxylin-
eosin stain; original magnification, ⫻50.)
liver parenchyma. The lesions had attenu-
ation similar to that of the aorta, but they
exhibited a heterogeneous appearance be-
cause of the presence of focal areas of hy-
perattenuation. In two patients with a his-
tory of Thorotrast exposure, multiple
lesions displaced the linear network of re-
sidual Thorotrast, and the spleen was atro-
phic and hyperattenuating due to prior
Thorotrast exposure.
Contrast-enhanced CT images (n ⫽ 8)
showed multiple tumor nodules in five
patients. In these patients, tumor nod-
ules were hypoattenuating. In three pa-
tients, most lesions contained focal areas
of enhancement (Fig 1a). The attenua-
tion of many foci of enhancement was
less than that of the aorta but greater
than that of the hepatic parenchyma. A
few foci of enhancement were isointense
with the aorta. Some areas of enhance-
ment had irregular shapes, and others
were ring shaped. Focal regions of en-
Volume 222 䡠 Number 3
hancement were located predominantly
in the central portion of each lesion, with
a few located peripherally. Two of these
three patients had innumerable nodular
masses.
On contrast-enhanced CT images, three
patients had a large, dominant mass lesion
in the liver. Two of these dominant masses
demonstrated heterogeneous enhance-
ment that suggested central necrosis and
fibrotic change (Fig 2). One of these pa-
tients had a solitary tumor that arose in
the vicinity of a chronic organized hema-
toma, and portal- and delayed-phase im-
aging demonstrated heterogeneous and
progressive enhancement of the tumor
(Fig 2). The second patient had a domi-
nant mass with multiple satellite nodules
in the liver. The third patient had two
dominant masses of homogeneous hy-
poattenuation without focal enhance-
ment.
On MR images, four patients had dom-
inant mass lesions, and three were ac-
companied by satellite lesions. On T1-
weighted images, all of the dominant
masses were of low intensity but con-
tained focal areas of high intensity, sug-
gesting hemorrhage (Figs 2– 4). On T2-
weighted images, these dominant masses
demonstrated predominantly increased
T2 signal intensity compared with that of
the skeletal muscle and the normal liver,
and the images showed a heterogeneous
appearance with focal areas of high in-
tensity interspersed with septumlike re-
gions of low intensity (Figs 2– 4). On T2-
weighted images, satellite lesions showed
high intensity (Fig 3) or fluid-fluid levels,
which are typical of intratumoral hemor-
rhage (Fig 4). One patient did not have a
focal intrahepatic mass, but rather had
diffuse heterogeneous signal intensity
throughout the liver on T1-weighted im-
ages (Fig 5). Corresponding T2-weighted
images showed no apparent mass lesion
(Fig 5).
On dynamic contrast-enhanced MR
images, the pattern of enhancement in
the arterial and portal phases was mark-
edly heterogeneous within the dominant
masses. Delayed imaging resulted in pro-
gressive enhancement over time com-
pared with the enhancement of the sur-
rounding hepatic parenchyma, except in
central regions of hemorrhage, necrosis,
or fibrosis.
Pathologic Findings
All tumors were characterized micro-
scopically by spindle-shaped cells, which
demonstrated vascular formation in var-
ious patterns. In specimens of substantial
size, necrosis was present in 10 of 12
(83%) specimens, and hemorrhage was
present in nine of 11 (82%) specimens
(one was too small to evaluate). The pres-
ence of necrosis and hemorrhage in the
examined samples correlated with high
T1-weighted signal intensity at MR imag-
ing in four of five tumors. Heterogeneous
attenuation on unenhanced CT scans
correlated with the presence of necrosis
in four of six patients and hemorrhage in
three of five patients. Cavernous spaces
lined with neoplastic cells were present
in six of eight (75%) cases with suffi-
ciently large sample volume. For exam-
ple, one patient with hemochromato-
sis and a dominant mass with satellite
tumors at imaging had a tumor that
contained numerous, blood-filled cystic
spaces (Fig 3). In another patient with a
multifocal tumor, images showed abun-
dant cavernous spaces lined with tumor
cells (Fig 1). These cavernous vascular
spaces simulated those seen with cavern-
ous hemangioma; interestingly, these tu-
mors demonstrated focal enhancement
at contrast-enhanced CT (Fig 1). Another
lesion, which arose at the site of a
chronic hematoma, was predominantly
solid and demonstrated neoplastic cells
(forming freely anastomosing vascular
channels with few dilated vascular spac-
es), central fibrosis, and hyalinization
(Fig 2).
Some tumors did have unique features.
MR images of one patient showed little
heterogeneity of signal intensity in the
liver. This patient had undergone splenec-
tomy and wedge resection in the lateral
segment 10 months prior to the imaging
study. At the time of resection, the surgeon
noted that the liver “contained a diffuse
process, appearing vascular in nature, char-
acterized by small, less-than-1-cm, reddish-
purple spots.” At histologic analysis, the
spleen and entire wedge resection were
diffusely involved with angiosarcoma,
with hyperchromatic tumor cells dif-
fusely spreading along the portal tracts
and sinusoidal spaces without forming
macronodular lesions (Fig 5).
DISCUSSION
Angiosarcoma is the most common pri-
mary sarcoma in the liver (1– 4). It com-
monly affects patients 60 –70 years of age
but is also known to occur in younger
patients (2,3). A strong male predomi-
nance has been reported, with a male-
female ratio of four to one (3).
Angiosarcoma has received attention
in recent decades because of its frequent
Primary Hepatic Angiosarcoma 䡠 669

association with several environmental
carcinogens, such as Thorotrast, vinyl
chloride, and arsenic compounds (5,6).
Both hemochromatosis and anabolic ste-
roids have been associated with angiosar-
coma (14,15), as seen in one patient with
aplastic anemia in our series. Chronic in-
flammation has also been implicated as a
potential cause of the development of
pleural angiosarcoma in patients with
pyothorax (16). In one of our patients,
angiosarcoma originated in or adjacent
to a chronic organized hematoma. How-
ever, in most patients with primary he-
patic angiosarcoma, cause is not appar-
ent (2,3).
Patients usually present with nonspe-
cific symptoms, including abdominal pain,
weakness, fatigue, anorexia, and weight
loss. Hematologic abnormalities, including
microangiopathic hemolytic anemia and
670 䡠 Radiology 䡠 March 2002
thrombocytopenia, are frequently associ-
ated (3). Most patients in our series were
anemic (62%) and thrombocytopenic
(54%), with microangiopathic hemolytic
anemia in 23%. Erythrocytes are thought
to be traumatized in the poorly organized
neoplastic vessels, and platelets may be
trapped. Occasionally, this localized co-
agulopathy results in the development of
systemic disseminated intravascular co-
agulation, which was seen in 31% of our
patients.
Most patients have metastatic lesions
at the time of presentation. The most
common site of metastases is the lung,
followed by the spleen (3). In our series,
23% of patients had pulmonary metasta-
ses, but 46% had splenic metastases. The
high frequency of splenic metastases is
notable, but in some instances it is diffi-
cult to determine whether the liver or the
Figure 2. Images show angiosarcoma associ-
ated with a chronic organized subcapsular he-
matoma (arrowheads) in a 76-year-old man.
(a) Transverse CT scan in the portal phase dem-
onstrates a heterogeneous enhancement pat-
tern in the lesion (arrows). At resection, the tu-
mor was found adherent to (but not invading)
the diaphragm. (b) Transverse contrast-en-
hanced dynamic delayed-phase CT scan dem-
onstrates progressive enhancement over time
(arrows). (c) Transverse T1-weighted spin-echo
(266/14) MR image shows a massive tumor (ar-
row) in the vicinity of a chronic organized sub-
capsular hematoma (arrowheads). The lesion
contains focal areas of high intensity, which
suggest hemorrhage. (d) Transverse fat-satu-
rated T2-weighted fast spin-echo (8,571/70) MR
image shows the marked heterogeneous appear-
ance of the lesion (arrows) and hematoma (ar-
rowheads). (e) Cut section of the gross specimen
shows a hemorrhagic mass with a chronic orga-
nized subcapsular hematoma (white arrow-
heads). Fibrotic scar (hyalinization), which is
whitish in color, is notable in the central area of
the lesion (black arrowhead).
spleen is the site of the primary neo-
plasm. It may be speculated that splenic
lesions represent simultaneous occur-
rence, which may be related to known or
unknown environmental carcinogens.
Diagnosis by means of liver biopsy has
been reported as treacherous and nondi-
agnostic (17). Our experience differs in
that seven of nine (78%) percutaneous
biopsies in our series yielded diagnostic
specimens without substantial bleeding.
Nevertheless, recent reports of lethal com-
plications following image-guided percuta-
neous biopsy remain (17). Therefore, surgi-
cal backup may be prudent if percutaneous
biopsy is attempted.
The gross appearance of angiosarcoma
at pathologic evaluation is characterized
Koyama et al
 findings of angiosarcoma. In the case of
Thorotrast exposure, which is well docu-
mented, tumors are hypoattenuating on
unenhanced CT scans and are inferred
by the displacement of the hyperat-
tenuating linear network of residual
Thorotrast (4 – 6). Even in patients with-
out Thorotrast exposure, unenhanced
CT images demonstrate the tumor to be
predominantly hypoattenuating com-
pared with the surrounding hepatic pa-
renchyma. On contrast-enhanced im-
ages, most lesions are hypoattenuating
compared with normal liver tissue, but
some lesions can be hyperattenuating
(8). While White et al have reported pro-
gressive centripetal enhancement similar
to that of cavernous hemangioma (10),
recent reports have demonstrated that
angiosarcoma does not resemble benign
Figure 3. MR images show angiosarcoma in a 37-year-old man with known aplastic anemia. cavernous hemangioma at biphasic im-
(a) Transverse T1-weighted spin-echo (250/17) MR image shows a diffuse lesion involving almost
the entire posterior segment of the right lobe of the liver. Numerous small nodules of high
aging (8,9,12). Peterson et al found only
intensity suggest a focal area of hemorrhage. Diffuse decrease in signal intensity in the liver, one tumor nodule in six patients that
spleen (not shown), and bone marrow is consistent with patient’s known secondary hemochro- had the typical nodular enhancement of
matosis. (b) Transverse fat-saturated T2-weighted spin-echo (2,500/120) MR image shows com- a hemangioma (8). We observed focal ar-
partmentalization within the lesion that contains numerous focal areas of high intensity. eas of enhancement in three of five (60%)
patients with multiple tumor nodules.
These lesions were different from multi-
ple cavernous hemangiomas for several
reasons: Focal areas of enhancement
showed less attenuation than the aorta,
with bizarre shapes, central enhance-
ment, or peripheral ring-shaped en-
hancement. Innumerable hepatic lesions
indicated that a malignancy was present.
Unlike the patient group of Peterson et
al, three patients in our series had large
dominant masses. None of these patients
demonstrated foci of enhancement simi-
lar to those of hemangioma.
Although there have been several re-
ports on the CT appearances of angiosar-
coma, little has been discussed on the MR
imaging features of angiosarcoma and
the role of MR imaging in preoperative
diagnosis. We found that MR imaging
Figure 4. MR images show angiosarcoma in a 65-year-old man. (a) Transverse T1-weighted demonstrated the hemorrhagic, hetero-
spin-echo (250/15) MR image shows multiple low-intensity lesions (arrows) that contain focal geneous, and hypervascular nature of all
areas of slightly high T1-weighted signal intensity. (b) Transverse fat-saturated T2-weighted the dominant masses in our series.
spin-echo (2,500/100) MR image shows heterogeneous signal intensity throughout the dominant On T1-weighted MR images, all four
mass. Fluid-fluid levels can be seen in smaller satellite lesions (arrows). dominant mass lesions contained irreg-
ular areas of high signal intensity, sug-
gesting hemorrhage. The high frequency
by the presence of remarkable necrosis lated sinusoidal or cavernous spaces that of hemorrhage in the lesions demon-
and hemorrhage and can show four types simulate cavernous hemangioma to slit- strated on T1-weighted images is in good
of growth patterns: multiple nodules, a like, freely anastomosing vascular chan- accordance with gross pathologic fea-
large dominant mass, mixed patterns of nels (2,18,19). These various vascular pat- tures. The presence of fluid-fluid levels on
a dominant mass with nodules, and, terns are usually intermingled with each T2-weighted images is another finding
rarely, a diffusely infiltrating micronodu- other, with the predominant pattern dif- that reflects the hemorrhagic nature of
lar tumor (18,19). Angiosarcoma is char- fering in each patient. Fibrosis and dep- angiosarcoma, but this finding is also of-
acterized microscopically by spindle- osition of hemosiderin are frequently en- ten seen with hypervascular metastases.
shaped cells that form vascular channels. countered in solid portions of the tumor All four dominant mass lesions that
These channels show a wide spectrum of (18,19). underwent T2-weighted MR imaging also
patterns, ranging from excessively di- There have been several reports on CT showed a markedly heterogeneous archi-

Volume 222 䡠 Number 3 Primary Hepatic Angiosarcoma 䡠 671

tecture, with focal areas of high intensity
along with septumlike or rounded areas
of low intensity on T2-weighted images.
This appearance suggests compartmen-
talization within the tumor, and it is sim-
ilar to that observed with hepatocellular
carcinoma. Areas of low signal intensity
on T2-weighted images may reflect he-
mosiderin, fibrous solid portions, or fresh
hemorrhage, on occasion, while areas of
high intensity may represent hemor-
rhage or necrosis. Unlike results in earlier
case reports, no tumors in our series
showed a central area of low intensity or
a few internal septa that could have been
confused with hemangioma (11,20,21).
On dynamic contrast-enhanced MR
images, each lesion showed heteroge-
neous enhancement on the arterial- and
portal-phase images. At delayed imaging,
however, there was progressive enhance-
ment of the lesion compared with that of
early-phase images. The heterogeneous
enhancement pattern on contrast-en-
hanced images was similar to that ob-
served with CT and likely represents the
heterogeneity of microscopic vascular
patterns within each tumor. The areas
with abundant, freely anastomosing vas-
cular channels may enhance quickly,
while dilated cavernous vascular spaces
may show slowly progressive enhance-
ment.
In one patient with diffuse signal het-
erogeneity throughout the liver on T1-
weighted MR images, pathologic findings Figure 5. Images show angiosarcoma that diffusely involves the entire liver in a 64-year-old-
were surprising, as tumor cells diffusely man. (a) Transverse T1-weighted spin-echo (500/15) MR image shows slightly heterogeneous
signal intensity of the liver and ascites. (b) Transverse T2-weighted spin-echo (2,000/100) MR
infiltrated the liver parenchyma along
image shows little abnormality. (c) Photomicrograph with a low-power view (hematoxylin-eosin
the portal tracts. This micronodular pat- stain; original magnification, ⫻25) shows hyperchromatic cells diffusely involving the liver
tern is an unusual feature of angiosar- parenchyma along the portal tracts without forming macronodular lesions.
coma (18).
Since angiosarcoma commonly ap-
pears as multiple masses or as a heteroge-
neous dominant mass, it often cannot be abnormalities, including thrombocytope- MR images can demonstrate its hemor-
readily distinguished from hypervascular nia, disseminated intravascular coagula- rhagic and heterogeneous appearance.
metastases (such as neuroendocrine tu- tion, and hemolytic anemia, are more fre- Dynamic enhancement of a dominant
mors) and hepatocellular carcinoma. All quently associated with angiosarcoma. mass at CT or MR imaging may show
of these tumors may demonstrate in- Our study had several weaknesses— heterogeneous enhancement on the early-
ternal hemorrhage and heterogeneity, in predominantly, small sample size and phase images and progressive enhance-
addition to early and heterogeneous lack of consistent imaging parameters, ment on delayed images. T2-weighted
enhancement (22,23). In contrast to hep- due to the rarity of the tumor and the MR images demonstrate a correspond-
atocellular carcinoma, however, angio- length of time in which cases were col- ing heterogeneous or compartmental-
sarcoma demonstrates continuing, pro- lected. Not all patients received intrave- ized appearance of the dominant mass.
gressive enhancement on delayed-phase nous contrast material or underwent When angiosarcoma appears as multi-
images. Splenic metastases and lack of dynamic imaging. Only two patients ple nodular lesions at CT, most lesions
cirrhosis may also suggest angiosarcoma underwent complete surgical resection, are hypoattenuating, and foci of en-
instead of hepatocellular carcinoma. which prohibited point-to-point radio- hancement may be present, but they are
Clinical findings may also help to distin- logic-pathologic correlation. Neverthe- clearly distinguishable from the nodular
guish angiosarcoma from other hypervas- less, we believe that some important enhancement in benign hemangiomas.
cular lesions. Hepatocellular carcinoma is and consistent patterns emerged. Such nodular enhancement may be less
usually associated with chronic hepatitis or Angiosarcoma has various appearances than that of the aorta. It is often bizarre
liver cirrhosis caused by chronic virus in- at CT or MR imaging that reflect its var- in shape, with ring enhancement seen in
fections, and it may be associated with el- ied histologic composition. When an an- some nodules. Splenic metastases are also
evated ␣-fetoprotein levels. Hematologic giosarcoma appears as a massive lesion, common with angiosarcoma. In a patient

672 䡠 Radiology 䡠 March 2002 Koyama et al

with a hypervascular and hemorrhagic
dominant mass or multiple liver lesions,
characteristic hematologic abnormalities
may suggest a diagnosis of primary he-
patic angiosarcoma.
Acknowledgments: We thank professors Yuji
Itai and David H. Stephens for their thoughtful
mentorship and encouragement of young radi-
ologists, and for introducing the authors. We
also thank Dr Itai for stimulating our interest in
primary hepatic angiosarcoma with his insight-
ful observations.
References
1. Alrenga DP. Primary angiosarcoma of the
liver: review article. Int Surg 1975; 60:
198 –203.
2. Ishak KG. Mesenchymal tumor of the
liver. In: Okuda K, Peters RL, eds. Hepa-
tocellular carcinoma. New York, NY:
Wiley, 1976; 247–308.
3. Locker GY, Doroshow JH, Zwelling LA,
Chabner BA. The clinical features of he-
patic angiosarcoma: a report of four cases
and a review of the English literature.
Medicine 1979; 58:48 – 64.
4. Buetow PC, Buck JL, Ros PR, Goodman
ZD. Malignant vascular tumors of the liv-
er: radiologic-pathologic correlation. Ra-
dioGraphics 1994; 14:153–156.
5. Levy DW, Rindsberg S, Friedman AC, et al.
Thorotrast-induced hepatosplenic neopla-
sia: CT identification. AJR Am J Roentgenol
1986; 146:997–1004.
6. Silverman PM, Ram PC, Korobkin M. CT
appearance of abdominal thorotrast dep-
Volume 222 䡠 Number 3
osition and Thorotrast-induced angiosar-
coma of the liver. J Comput Assist To-
mogr 1983; 7:655– 658.
7. Vasile N, Larde D, Zafrani ES, Berard H,
Mathieu D. Hepatic angiosarcoma. J Com-
put Assist Tomogr 1983; 7:899–901.
8. Peterson MS, Baron RL, Rankin SC. He-
patic angiosarcoma: findings on mul-
tiphasic contrast-enhanced helical CT do
not mimic hepatic hemangioma. AJR
Am J Roentgenol 2000; 175:165–170.
9. Rademaker J, Widjaja A, Galanski M. He-
patic hemangiosarcoma: imaging find-
ings and differential diagnosis. Eur Radiol
2000; 10:129 –133.
10. White PG, Adams H, Smith PM. The com-
puted tomographic appearances of angio-
sarcoma of the liver. Clin Radiol 1993;
48:321–325.
11. Itai Y, Teraoka T. Angiosarcoma of the
liver mimicking cavernous hemangioma
on dynamic CT. J Comput Assist Tomogr
1989; 13:910 –912.
12. Ohtomo K, Araki T, Itai Y, et al. MR im-
aging of malignant mesenchymal tumors
of the liver. Gastrointest Radiol 1992; 17:
58 – 62.
13. Worawattanakul S, Semelka RC, Kelekis
NL, Woosley JT. Angiosarcoma of the liv-
er: MR imaging pre- and post-chemother-
apy. Magn Reson Imaging 1997; 15:613–
617.
14. Falk H, Thomas LB, Popper H, Ishak KG.
Hepatic angiosarcoma associated with
androgenic-anabolic steroids. Lancet 1979;
2:1120–1123.
15. Fracanzani AL, Conte D, Fraquelli M, et
al. Increased cancer risk in a cohort of 230
patients with hereditary hemochromato-
sis in comparison to matched control pa-
tients with non-iron-related chronic liver
disease. Hepatology 2001; 33:647– 651.
16. Myoui A, Aozasa K, Iuchi K, et al. Soft
tissue sarcoma of the pleural cavity. Can-
cer 1991; 68:1550 –1554.
17. Drinkovic I, Brkljacic B. Two cases of le-
thal complications following ultrasound-
guided percutaneous fine-needle biopsy
of the liver. Cardiovasc Intervent Radiol
1996; 19:360 –363.
18. Kojiro M, Nakashima T, Ito Y, Ikezaki H,
Mori T, Kido C. Thorium dioxide-related
angiosarcoma of the liver: pathomorpho-
logic study of 29 autopsy cases. Arch
Pathol Lab Med 1985; 109:853– 857.
19. Ludwig J, Hoffman HN. Hemangiosar-
coma of the liver. Spectrum of morpho-
logic changes and clinical findings. Mayo
Clin Proc 1975; 50:255–263.
20. Ohtomo K, Araki T, Itai Y, et al. MR im-
aging of malignant mesenchymal tumors
of the liver. Gastrointest Radiol 1992; 17:
58 – 62.
21. Ros PR, Lubbers PR, Olmsted WW, Mo-
rillo G. Hemangioma of the liver: hetero-
geneous appearance on T2-weighted im-
ages. AJR Am J Roentgenol 1987; 149:
1167–1170.
22. Onaya H, Itai Y. MR imaging of hepato-
cellular carcinoma. Magn Reson Imaging
Clin N Am 2000; 8:757–768.
23. Stevens WR, Gulino SP, Batts KP, Ste-
phens DH, Johnson CD. Mosaic pattern
of hepatocellular carcinoma: histologic
basis for a characteristic CT appearance.
J Comput Assist Tomogr 1996; 20:337–
342.
Primary Hepatic Angiosarcoma 䡠 673