Gastrointestinal Imaging • Original Research

Choi et al.
MRI of Small HCC

Gastrointestinal Imaging
Original Research

MRI of Small Hepatocellular

Downloaded from www.ajronline.org by 125.161.136.52 on 04/18/18 from IP address 125.161.136.52. Copyright ARRS. For personal use only; all rights reserved

Carcinoma: Typical Features Are

Less Frequent Below a Size Cutoff
of 1.5 cm
Moon Hyung Choi1 OBJECTIVE. The purposes of this study were to analyze MRI features of small hepato-
Joon-Il Choi1 cellular carcinomas (HCCs) on the basis of size and to evaluate the difference in frequency
Young Joon Lee1 of typical radiologic hallmarks of HCC (arterial enhancement and washout) according to the
Michael Yong Park1 tumor size.
Sung Eun Rha1 MATERIALS AND METHODS. Enrolled were 86 patients with 110 HCCs 3 cm or
smaller who underwent surgical resection or transplantation. Two radiologists reviewed gadox-
Chandana Lall2
etic acid–enhanced MRI features for signal intensity of T2-weighted and T1-weighted imaging,
Choi MH, Choi JI, Lee YJ, et al. diffusion restriction, presence of arterial enhancement, washout on portal and transitional phas-
es, and signal intensity on the hepatobiliary phase. ROC curve analysis was performed to deter-
mine the optimal HCC cutoff size for radiologic hallmarks of HCC. Tumors were divided into
two groups by cutoff size, and the frequencies of MRI features were assessed.
RESULTS. On ROC analysis, the optimal cutoff for radiologic hallmarks of HCC was
1.5 cm in independent and consensus reviews by two radiologists. HCCs smaller than 1.5 cm
showed typical finding of HCC less frequently than HCCs 1.5 cm or larger in diameter. In
subgroup analyses, HCCs with diameters between 1 and 1.5 cm showed similar MRI findings
to HCCs with diameters 1 cm or less but significantly different findings compared with HCCs
with diameters from 1.5 to 2 cm and 2–3 cm.
CONCLUSION. HCCs smaller than 1.5 cm in size less frequently showed MRI findings
seen typically in larger HCCs. Therefore, small HCCs are harder to detect with certainty not
only because of small size but also because of the lower frequency of typical MRI findings.

Keywords: diagnostic criteria, gadoxetic acid, hepatic
carcinoma, MRI
DOI:10.2214/AJR.16.16414
Received March 6, 2016; accepted after revision
September 28, 2016.
This study was supported by grant 150160 from the
National R&D Program for Cancer Control, Ministry of
Health & Welfare, Republic of Korea.
1
Department of Radiology and Cancer Research Institute,
Seoul St. Mary’s Hospital, College of Medicine, The
Catholic University of Korea, 222, Banpo-daero,
Seocho-gu, Seoul, 137-701, Republic of Korea. Address
correspondence to J. I. Choi (dumkycji@gmail.com).
2
Department of Radiological Sciences, University of
California, Irvine, Orange, CA.
AJR 2017; 208:544–551
0361–803X/17/2083–544
© American Roentgen Ray Society
nlike common malignancies such
U as lung and colon cancer, defini-
tive diagnosis of hepatocellular
carcinoma (HCC) can be made
noninvasively and without histopathologic
confirmation on the basis of imaging findings
of a cirrhotic liver [1–4]. Well-known unique
HCC imaging features include intense hyper-
vascularity on the arterial phase followed by
contrast washout in the portal venous and de-
layed phases. The specificity for these HCC
radiologic hallmarks is 90–95% [1, 2, 5–7].
Well-established guidelines for managing
HCC use these radiologic hallmarks as defi-
nite findings for noninvasive imaging diagno-
sis using dynamic CT or MRI [1–4, 8].
However, organizations disagree about the
details of diagnostic criteria in guidelines for
diagnosis of small HCCs, such as cutoff sizes
for radiologic examinations. The American
Association for the Study of Liver Diseas-
es (AASLD) recommends diagnostic crite-
ria of arterial enhancement and washout for
all nodules larger than 1 cm. The European
Association for the Study of the Liver–Eu-
ropean Organisation for Research and Treat-
ment of Cancer (EASL-EORTC) guidelines
divide nodules into three groups by size (< 1
cm,  ≥ 1 cm but  < 2 cm,  ≥ 2 cm) and apply
different diagnostic criteria to the groups.
The Liver Imaging Reporting and Data Sys-
tem (LI-RADS) divides nodules by a cut-
off diameter of 2 cm [9]. Furthermore, some
pathologists have arbitrarily defined small
HCC as carcinoma that measures less than 2
cm in diameter [10]. However, the guideline
cutoff sizes of 1 or 2 cm for different diag-
nostic criteria or the pathologic definition of
small HCC have little scientific basis. We be-
lieve the cutoffs are based on common con-
vention or clinical convenience.
We hypothesized that smaller HCCs show
typical imaging features of HCC less fre-
quently. The aims of this study were to ana-
lyze MRI features of small HCCs (≤ 3 cm)
on the basis of size and to evaluate the differ-

544 AJR:208, March 2017

 MRI of Small HCC

ence in frequency of typical radiologic hall- TABLE 1:  MRI Sequence Parameters
marks of HCC (arterial enhancement and
Section Flip Matrix
washout) according to the tumor size. Sequence Thickness (mm) TR/TE Angle (°) Size

Materials and Methods T2-weighted FSE with fat saturation 6 4400/100 160 448 × 202
Patient Selection T2-weighted HASTE 6 900.0/95 138 320 × 144
Downloaded from www.ajronline.org by 125.161.136.52 on 04/18/18 from IP address 125.161.136.52. Copyright ARRS. For personal use only; all rights reserved

This study was approved by Seoul St. Mary's
Hospital's institutional review board. The informed
consent requirement was waived because data col-
lection was retrospective. We reviewed the records
of all patients with pathologically confirmed HCC
who underwent any type of hepatic resection in-
cluding wedge resection, segmentectomy, sectio-
nectomy, hemihepatectomy, and total hepatectomy
for transplantation from January 2011 to May 2014
at our institution. The study included only patients
who underwent gadoxetic acid–enhanced MRI
within 30 days of surgery to avoid including pa-
tients who developed HCC between the MRI exam-
ination and surgical resection. Among 210 patients
who underwent hepatic resection and underwent
preoperative MRI, 114 had HCCs that were 3 cm
or smaller. Of these 114 patients, 28 were exclud-
ed because of prior treatment with chemoemboliza-
tion or radiofrequency ablation (n = 20); lesion loca-
tion not in the pathologic report (n = 3); more than
10 HCCs (n = 2); pathologically confirmed HCCs
that were not visible on MRI even with comparison
with a gross image of the resected liver (n = 2); and
poor-quality MRI because of ascites (n = 1). Includ-
ed were 110 pathologically proven HCCs of 3 cm or
larger in 86 patients. The 86 patients consisted of 64
men and 22 women with a mean age of 58.9 years
(range, 30–77 years). Surgical procedures included
total hepatectomy in 24 patients, hemihepatectomy
in 15 patients, sectionectomy in seven patients, seg-
mentectomy in 10 patients, and wedge resection in
30 patients. The mean time between MRI and sur-
gery was 18.3 days (range, 1–30 days).
We also analyzed the tumor differentiation on
pathologic examinations according to the Edmond-
son-Steiner grade. We classified HCCs into two
groups: well-differentiated HCCs (Edmondson-
Steiner grade I) and moderately or poorly differ-
entiated HCCs (Edmondson-Steiner grades II–IV).
MRI
A 3-T MRI scanner (Verio, Siemens Healthcare)
at our institution with an 8-channel phased-array
torso coil was used. We performed unenhanced fast
spin-echo and fat-suppressed single-shot fast spin-
echo T2-weighted imaging (fast spin-echo with fat
saturation and HASTE) and T1-weighted imaging
with in-phase and opposed-phase dual gradient echo.
For contrast-enhanced dynamic imaging, 0.025
mmol/kg of gadoxetic acid (Primovist or Eovist,
Bayer Schering Pharma) was injected at a rate of
2 mL/s by an automated infusion system followed
In-phase and opposed-phase 6 130/2.6 or 1.3 52 288 × 173
T1-weighted GRE 2 3.1/1.1 11.5 360 × 270
DWI 6 3500/47 90 140 × 96
Note—FSE = fast spin-echo, GRE = gradient-recalled echo.
by a 20-mL saline chaser. Three-dimensional gra- ber, size, and location of lesions and retrospective-
dient-recalled echo imaging was performed in the ly reviewed gadoxetic acid–enhanced MRI stud-
arterial phase (30- to 35-second delay with the bo- ies performed before surgery. Recorded imaging
lus-tracking technique), portal venous phase (65- features of lesions included tumor signal intensi-
to 80-second delay), transitional phase (180-sec- ty on T1-weighted imaging and T2-weighted im-
ond delay), and hepatobiliary phase (20-minute aging, presence of arterial enhancement and por-
delay). We performed DWI with echo-planar imag- tal or transitional phase washout, signal intensity
ing applying b values of 0, 50, 500, and 800 s/mm2. on the hepatobiliary phase, presence of a fat com-
MRI sequence parameters are shown in Table 1. ponent, presence of rim enhancement, and pres-
ence of diffusion restriction. Arterial enhance-
Image Analysis ment was defined as hyperintensity on the arterial
An abdominal radiologist with 6 years of expe- phase compared with surrounding liver parenchy-
rience interpreting liver MRI reviewed all preop- ma in unenhanced hypointense or isointense le-
erative MRI studies and measured maximal diam- sions. Portal or transitional washout was defined
eter of lesions on axial images, correlating tumor as nodule hypointensity on the portal or transition-
location with pathology report descriptions. The al phase compared with surrounding liver paren-
hepatobiliary phase was primarily used for diam- chyma. The washout definition was from the 2014
eter measurements. The diameter of lesions not Korean practice guidelines for HCC manage-
visible on the hepatobiliary phase was measured ment [8]. Signal intensity drop on opposed-phase
using T2-weighted imaging. For lesions not dis- T1-weighted imaging indicated the presence of a
cernible on T2-weighted imaging, arterial phase fat component in the nodule. A hyperintense rim
imaging was used. around the nodule on the portal or transitional
Two abdominal radiologists with 19 and 6 years phase was considered ringlike enhancement. Dif-
of experience read examinations in consensus and fusion restriction was high signal intensity on DWI
independently with more than a 2-month interval with a b value of 800 s/mm2 and a lower appar-
to remove recall bias. Both were aware of the num- ent diffusion coefficient of the nodule compared
TABLE 2: Interreader Agreement of Imaging Findings of Hepatocellular
Carcinoma (HCC)
MRI Features Reader 1 vs Reader 2 Reader 1 vs Consensus Reader 2 vs Consensus
T2 hyperintensity 0.669 0.856 0.720
T1 hypointensity 0.734 0.898 0.749
Arterial enhancement 0.769 0.812 0.722
Portal or transitional washout 0.796 0.813 0.667
Hypointensity on hepatobiliary 0.740 0.756 0.782
phase
Diffusion restriction 0.681 0.872 0.724
Ringlike enhancement 0.536 0.854 0.470
Fat component 0.660 0.798 0.526
Radiologic hallmarka 0.734 0.898 0.707
Note—Data are kappa statistics evaluated by two readers independently and in consensus.
aRadiologic hallmark = the radiologic hallmark of HCC according to the American Association for the Study of
Liver Diseases or European Association for the Study of the Liver–European Organisation for Research and
Treatment of Cancer guidelines (arterial enhancement and washout).

AJR:208, March 2017 545

 Choi et al.

1.0 1.0 1.0

1.5 cm 1.5 cm 1.5 cm

0.8 0.8 0.8
Sensitivity

Sensitivity

Sensitivity
0.6 0.6 0.6
Downloaded from www.ajronline.org by 125.161.136.52 on 04/18/18 from IP address 125.161.136.52. Copyright ARRS. For personal use only; all rights reserved

0.4 0.4 0.4

0.2 0.2 0.2

0.0 0.0 0.0

0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0
1 – Specificity 1 – Specificity 1 – Specificity

A B C
Fig. 1—ROC curve analyses show radiologic hallmarks for noninvasive diagnosis of hepatocellular carcinoma.
A, In consensus review, area under ROC curve was 0.721.
B, In review by radiologist 1, area under ROC curve was 0.712.
C, In review by radiologist 2, area under ROC curve was 0.736. Optimal cutoff was 1.5 cm in all three curves.

with liver parenchyma. All nodules were assessed
whether or not lesions showed HCC radiolog-
ic hallmarks of hypervascularity on the arterial
phase and washout on the portal or transitional
phase. Because features other than MRI radiolog-
ic hallmarks are helpful for diagnosing HCC, we
evaluated hyperintensity on T2-weighted imaging,
hypointensity on the hepatobiliary phase, and dif-
fusion restriction. These additional features are
defined in LI-RADS as ancillary features that may
favor malignancy [11].
Statistical Analysis
Interobserver agreement in interpreting images
between the two radiologists’ independent review
and in consensus review was evaluated by using kap-
pa statistics, and results were classified as follows:
poor, < 0; slight, 0–0.20; fair, 0.21–0.40; moderate,
0.41–0.60; good, 0.61–0.80; and excellent, 0.81–1.00.
ROC curve analysis was used to find the opti-
mal cutoff tumor size for HCC radiologic hallmarks.
ROC curve analysis was performed for the two radi-
ologists’ independent reading and in consensus read-
ing. We classified tumors into two groups by this
cutoff. The Fisher exact test was used to assess statis-
tical differences in frequencies of MRI features be-
tween smaller and larger tumor groups; p values less
than 0.05 were considered to indicate significant dif-
ferences. We performed t tests to compare tumor di-
ameters on pathologic examination and MRI.
For subgroup analysis, the two groups were
subdivided into four subgroups on the basis of ra-
diologically measured tumor diameter. The Fisher
exact test was used for subgroup analysis for MRI
feature frequencies. The results of consensus re-
view were used to perform subgroup analysis.
We also evaluated the difference of MRI fea-
tures between well-differentiated HCCs and mod-
erately or poorly differentiated HCCs. The Fisher
exact test was also used for subgroup analysis for
MRI feature frequencies.
Results MRI features and HCC hallmarks of T1 hy-
The average size of all HCCs was 1.78 ± 0.66 pointensity, T2 hyperintensity, hepatobili-
cm (range, 0.5–3.2 cm). No major discrepancy ary phase hypointensity, washout on portal or
was observed between tumor sizes measured transitional phase, intratumoral fat, rim en-
from MRI or obtained from histopathologic hancement, and diffusion restriction but not
reports (1.74 ± 0.74 cm; range, 0.4–3 cm) (p = arterial enhancement showed significant dif-
0.238). Fifteen HCCs were well-differentiated ferences in frequency between the smaller
and 95 were moderately or poorly differentiat- (< 1.5 cm) and larger (≥ 1.5 cm) tumor groups
ed. The kappa statistics showed good or excel- (Table 3). The larger the HCC, the more fre-
lent agreement among the three reviewers for quent the T2 hyperintensity, T1 hypointensi-
all imaging features except ringlike enhance- ty, portal or transitional washout, ringlike en-
ment (Table 2). Interreader agreement was bet- hancement, and radiologic hallmarks of HCC
ter between radiologist 1 and consensus re- were observed. Fat component was not visible
views than the other two sets of reviews. for all small tumors and was visible in 3.5% of
Analysis of the relationship between tumor the larger tumors. The radiologic hallmarks of
size and HCC radiologic hallmarks by ROC HCC were found in 79.7% (59/74) of the larger
curve suggested the optimal cutoff size was tumor group and 44.4% (16/36) of the smaller
1.5 cm in consensus and independent reviews tumor group (p < 0.001).
(Fig. 1). Tumors were divided into groups us- After analyzing the findings for the two
ing 1.5 cm as a cutoff, with 36 HCCs in the HCC size groups, we subdivided the tumors
smaller group and 74 in the larger group. All into four subgroups: subgroup 1, < 1 cm; sub-
TABLE 3:  MRI Features of Hepatocellular Carcinoma (HCC) in Two Size Groups
Smaller Tumor Group Larger Tumor Group
MRI Features (< 1.5 cm, n = 36) (≥ 1.5 cm, n = 74) pb
T2 hyperintensity 18 (50.0) 63 (85.1) < 0.001
T1 hypointensity 18 (50.0) 57 (77.0) 0.008
Arterial enhancement 29 (80.6) 66 (89.2) 0.244
Portal or transitional washout 19 (52.8) 61 (82.4) 0.003
Hypointensity on hepatobiliary phase 25 (69.4) 67 (90.5) 0.011
Diffusion restriction 15 (41.7) 60 (81.1) < 0.001
Ringlike enhancement 1 (2.8) 15 (20.3) 0.019
Fat component 0 (0.0) 10 (13.5) 0.029
Radiologic hallmarka 16 (44.4) 59 (79.7) < 0.001
Note—Data in parentheses are percentages.
aRadiologic hallmark = the radiologic hallmark of HCC according to the American Association for the Study of
Liver Diseases or European Association for the Study of the Liver–European Organisation for Research and
Treatment of Cancer guidelines (arterial enhancement and washout).
bThe p values were calculated using the Fisher exact test.

546 AJR:208, March 2017

 MRI of Small HCC
Downloaded from www.ajronline.org by 125.161.136.52 on 04/18/18 from IP address 125.161.136.52. Copyright ARRS. For personal use only; all rights reserved

A B C
Fig. 2—63-year-old man with 0.8-cm hepatocellular carcinoma (HCC) in segment VII of liver (subgroup 1).
A and B, Nodule shows subtle hyperintensity (arrow, A) on T2-weighted image (A) and enhancement (arrow, B)
on arterial phase image (B).
C, No washout is seen on transitional phase image.
D, DW image obtained at b value of 800 s/mm2 shows no restricted diffusion. This tumor did not show radiologic
hallmarks of HCC.

group 2, ≥ 1 to < 1.5 cm; subgroup 3, ≥ 1.5 to <
2 cm; and subgroup 4, ≥ 2 cm. There were 14
HCCs in subgroup 1, 22 HCCs in subgroup
2, 31 HCCs in subgroup 3, and 43 HCCs in
subgroup 4. Mean sizes in the four subgroups
were 0.74  ± 0.12 cm (range, 0.5–0.9 cm),
1.22 ± 0.12 cm (range, 1.0–1.4 cm), 1.71 ± 0.16
cm (range, 1.5–1.9 cm), and 2.45  ± 0.43 cm
(range, 2.0–3.2 cm), respectively.
No MRI features showed significant differ-
ences between subgroups 1 and 2. Four MRI
features were significantly more frequent in
subgroup 3 than subgroup 2: T2 hyperinten-
sity, portal or transitional washout, diffusion
restriction, and the radiologic hallmarks of
HCC. The radiologic hallmarks of HCC were
significantly more frequently observed in sub-
group 3 than subgroup 2 (p = 0.01). Compar-
ing subgroups 3 and 4, only T1 hypointensi-
ty showed a significantly different frequency.
MRI characteristics of subgroups and results
of the Fisher exact tests are shown in Table
4. Representative cases from each group are
shown in Figures 2–4.
Analysis according to HCC differentia-
tion showed that 15 HCCs were well differ-
entiated and 95 HCCs were moderately or
poorly differentiated. There were seven and
29 HCCs that were smaller than 1.5 cm in the
two differentiation groups, respectively (p =
0.244). Well-differentiated HCCs showed T1
hypointensity and fat more frequently than
did the moderately or poorly differentiated
HCCs (p = 0.017 and 0.029, respectively). Ex-
cept for these two MRI features, there were
no significant differences of MRI features be-
tween well-differentiated and moderately or
poorly differentiated HCCs (Table 5).
Discussion
In most guidelines, diagnostic criteria dif-
fer for small HCCs by cutoff sizes of 1 or

AJR:208, March 2017 547

 Choi et al.
Downloaded from www.ajronline.org by 125.161.136.52 on 04/18/18 from IP address 125.161.136.52. Copyright ARRS. For personal use only; all rights reserved

A B C
Fig. 3—70-year-old woman with 1.2-cm hepatocellular carcinoma (HCC) in segment VIII of liver (subgroup 2).
A, T2-weighted image shows lesion with subtle hyperintensity (arrow).
B and C, On dynamic contrast-enhanced image (B), tumor shows arterial enhancement (arrow, B) but no washout
(arrow, C) is seen on portal venous phase image (C).
D, Hepatobiliary phase image shows no-hypointensity lesion. This tumor did not show radiologic hallmarks of HCC.

2 cm. These cutoff sizes were chosen, accord-
ing to publications cited in the guidelines, as
arbitrary sizes for clinical convenience. No
scientific evidence supports using these sizes
as diagnostic criteria. Our study investigated
whether these cutoff sizes are appropriate.
We evaluated MRI features using gadoxet-
ic acid enhancement because MRI provides
more diverse imaging features for HCC than
CT. These features include signal intensity
on T1-weighted imaging, T2-weighted im-
aging, and the hepatobiliary phase and dif-
fusion restriction. Gadoxetic acid, a hepato-
biliary-specific contrast agent, is a standard
contrast agent for liver MRI in our institute
and showed robust performance for detect-
ing and characterizing HCC. When com-
bined with DWI, several studies found su-
perior diagnostic performance for gadoxetic
acid–enhanced MRI for diagnosing HCC
compared with using conventional gadolini-
um-based MRI or MDCT agents [12, 13].
In our study, the MRI features for HCC
including T2 hyperintensity, portal or tran-
sitional phase washout, and diffusion re-
striction as well as the radiologic hallmarks
of HCC were seen in most tumors 1.5 cm or
larger but in fewer than half of the tumors
smaller than 1.5 cm. The results of our sub-
group analysis reinforced the results for
a cutoff of 1.5 cm. Four imaging features
were different between subgroups 2 (≥ 1
cm to < 1.5 cm) and 3 (≥ 1.5 cm to < 2 cm).

548 AJR:208, March 2017

 MRI of Small HCC

TABLE 4:  Subgroup Analysis for MRI Features of Hepatocellular Carcinoma (HCC)
Subgroup 1 Subgroup 2 Subgroup 3 Subgroup 4 pb pb pb
MRI Features (n = 14) (n = 22) (n = 31) (n = 43) (Subgroup 1 vs 2) (Subgroup 2 vs 3) (Subgroup 3 vs 4)
T2 hyperintensity 7 (50) 11 (50) 25 (80.6) 38 (88.4) 1 0.035 0.51
T1 hypointensity 6 (42.9) 12 (54.5) 20 (64.5) 37 (86) 0.733 0.572 0.048
Downloaded from www.ajronline.org by 125.161.136.52 on 04/18/18 from IP address 125.161.136.52. Copyright ARRS. For personal use only; all rights reserved

Arterial enhancement 9 (64.3) 20 (90.9) 28 (90.3) 38 (88.4) 0.084 1 1

Portal or transitional washout 9 (64.3) 10 (45.5) 25 (80.6) 36 (83.7) 0.322 0.017 0.765
Hypointensity on hepatobiliary phase 12 (85.7) 13 (59.1) 26 (83.9) 41 (95.3) 0.142 0.061 0.122
Diffusion restriction 7 (50) 8 (36.4) 22 (71) 38 (88.4) 0.499 0.023 0.075
Ringlike enhancement 0 (0) 1 (4.5) 5 (16.1) 10 (23.3) 1 0.382 0.563
Fat component 0 (0) 0 (0) 4 (12.9) 6 (14) NA 0.132 1
Radiologic hallmarka 7 (50) 9 (40.9) 24 (77.4) 35 (81.4) 0.734 0.01 0.772
Note—Subgroup 1, < 1 cm; subgroup 2, ≥ 1 to < 1.5 cm; subgroup 3, ≥ 1.5 to < 2 cm; and subgroup 4, ≥ 2 cm. Data in parentheses are percentages. NA indicates not
applicable.
aRadiologic hallmark = the radiologic hallmark of HCC according to the American Association for the Study of Liver Diseases or European Association for the Study of the

Liver–European Organisation for Research and Treatment of Cancer guidelines (arterial enhancement and washout).
bThe p values were calculated using the Fisher exact test.

A B C

Fig. 4—51-year-old man with 1.7-cm hepatocellular carcinoma (HCC) in segment VI of liver (subgroup 3).
A, On T2-weighted image, this tumor shows T2 hyperintensity (arrow).
B, On arterial phase image, this tumor shows arterial enhancement (arrow).
C, On portal venous phase image, tumor shows washout (arrow).
D, On DW image obtained at b value of 800 s/mm2 , this tumor shows diffusion restriction (arrow). This tumor
shows radiologic hallmarks of HCC.
D

AJR:208, March 2017 549

 Choi et al.

Imaging characteristics of subgroup 2 were TABLE 5:  MRI Features According to Hepatocellular Carcinoma (HCC)
similar to subgroup 1 (< 1 cm). Subgroup 3 Differentiation by Edmonson-Steiner Grade
showed similar imaging features as subgroup Moderately or Poorly
4 (≥ 2 cm). This result suggests that dramatic Well-Differentiated HCC Differentiated HCC
changes in HCC may occur at 1.5 cm in di- MRI Features (n = 15) (n = 95) pb
ameter. Therefore, radiologists may diagnose Smaller than 1.5 cm 7 (46.7) 29 (30.5) 0.244
Downloaded from www.ajronline.org by 125.161.136.52 on 04/18/18 from IP address 125.161.136.52. Copyright ARRS. For personal use only; all rights reserved

a tumor smaller than 1.5 cm with less confi-

T2 hyperintensity 8 (53.3) 73 (76.8) 0.066
dence compared with larger tumors because
of the relative decreased frequency of see- T1 hypointensity 6 (40.0) 69 (72.6) 0.017
ing MRI findings of HCC in smaller HCCs. Arterial enhancement 12 (80.0) 83 (87.4) 0.428
Furthermore, because small lesions without Portal or transitional washout 8 (53.3) 72 (75.8) 0.114
typical findings of HCC could be HCCs, be-
nign lesions would be mistaken as HCCs and Hypointensity on hepatobiliary 13 (86.7) 79 (83.2) 1.000
phase
the likelihood of a small HCC being con-
fused with a benign lesion may be increased. Diffusion restriction 8 (53.3) 67 (70.5) 0.234
Typical and ancillary MRI findings of HCC Ringlike enhancement 3 (20.0) 13 (13.7) 0.455
should be carefully evaluated to differentiate Fat component 4 (26.7) 6 (6.3) 0.029
between small benign lesions and HCCs.
Radiologic hallmarka 8 (53.3) 67 (70.5) 0.234
No significant difference was observed
between tumors 1.5 cm or larger and tumors Note—Data in parentheses are percentages.
aRadiologic hallmark = the radiologic hallmark of HCC according to the American Association for the Study of
smaller than 1.5 cm in frequency of arterial Liver Diseases or European Association for the Study of the Liver–European Organisation for Research and
enhancement. However, portal or transitional Treatment of Cancer guidelines (arterial enhancement and washout).
phase washout was more frequently observed
bThe p values were calculated using the Fisher exact test.

in tumors 1.5 cm or larger than in smaller tu-

mors, which is why the smaller tumors less
frequently showed radiologic hallmarks of
HCC. This result agreed with prior studies
that found that HCCs less than 1 cm in size
frequently showed arterial enhancement [14],
and some small HCCs showed only arterial
enhancement without early washout [12, 15].
Therefore, when a small arterial nodule with-
out washout is newly detected in daily prac-
tice, other ancillary features are also consid-
ered to avoid misdiagnosing true HCC as a
benign arterial enhancing nodule such as an
arterioportal shunt.
The value of T2-weighted imaging for HCC
diagnosis is controversial [6, 16]. However, in-
creased signal intensity on T2-weighted im-
aging correlates with increased intratumor-
al arterial supply and decreased intratumoral
portal blood supply [17]. Another study found
that tumors with hypervascularity, high histo-
logic grade, and encapsulated growth are hy-
perintense on T2-weighted imaging [18]. In
our study, T2 hyperintensity was observed
more frequently as tumor size increased,
consistent with prior studies. In addition, the
significant difference between subgroups 2
and 3 in T2 hyperintensity frequency suggest-
ed that pathologic changes reflected T2 hyper-
intensity in tumors around 1.5 cm.
Hepatobiliary phase images may improve
HCC diagnosis [19] because 85.7% (12/14) of
nodules smaller than 1 cm showed low signal
intensity on the hepatobiliary phase in our
study, which was more frequent than wash-
out (52.8%). This result also agreed with an
earlier study that found that even well-differ-
entiated early HCC can be hypointense in the
hepatobiliary phase [15]. However, specifici-
ty is lost when low hepatobiliary phase signal
intensity is considered to be washout. Joo et
al. [20] and Choi et al. [21] reported 93.8%
sensitivity when hypointensity on hepatobili-
ary phase imaging was considered as wash-
out. Specificity was 48.4% and hypointensity
on hepatobiliary phase imaging was consid-
ered an ancillary feature, not washout. Our
study showed a higher frequency of diffusion
restriction in the larger tumor group than the
smaller tumor group. Our results are consis-
tent with a previous study showing that DWI
did not significantly improve detection of
HCCs of 2 cm or smaller [22].
Some studies support the division of small
HCCs into two clinicopathologic groups that
have been termed “early HCC” and “pro-
gressed HCC” [23]. Early HCC has a vaguely
nodular appearance and is well differentiated.
Progressed HCC has a distinctly nodular pat-
tern and is mostly moderately differentiated,
often with evidence of microvascular invasion.
Thus, we expected that most tumors small-
er than 1.5 cm may correspond to early HCC
and tumors of 1.5 cm or larger to progressed
HCC. However, our results showed that the in-
cidence of well-differentiated HCCs was not
significantly different between smaller tumors
and larger tumors with the cutoff size of 1.5
cm. Also, MRI features of well-differentiated
HCCs and moderately or poorly differentiated
HCCs were not different except for T1 signal
and the presence of fat. Therefore, the histo-
logic degree may be reflected for MRI features
in combination with the size of HCC. A study
with more patients is necessary to reveal the
impact of the combination of histologic degree
and size of HCC on MRI findings.
To overcome the low sensitivity of the
EASL-EORTC or AASLD criteria for diag-
nosing HCC, some studies have designed a
scoring system using additional MRI findings
[24]. Rhee et al. [24] reported that a combina-
tion of seven gadoxetic acid–enhanced MRI
features (arterial enhancement, washout, T1
hypointensity, T2 hyperintensity, hepatobili-
ary phase hypointensity, diffusion restriction,
and nodules  ≥ 1.5 cm) enhanced the sensi-
tivity of noninvasive diagnosis of early HCC
without sacrificing specificity compared with
the performance of the AASLD criteria. The
cutoff size for these diagnostic criteria is 1.5
cm. The LI-RADS system is another effort to-
ward this goal. In LI-RADS, the presence of
arterial enhancement, washout, capsule, and
threshold growth is important for diagnos-
ing HCC. However, ancillary features such as
T2 hyperintensity, diffusion restriction, tran-
sitional or hepatobiliary phase hypointensity,
and intralesional fat are considered imaging
features that modify the likelihood of HCC.
Our study has several limitations. First, the
retrospective design and review of only patho-
logically confirmed HCCs might have result-

550 AJR:208, March 2017


ed in selection bias. Except for patients who
underwent total hepatectomy for liver trans-
plantation, hepatic resection was performed
for nodules diagnosed radiologically. There-
fore, nodules with atypical radiologic findings
might not have been surgically resected and
Downloaded from www.ajronline.org by 125.161.136.52 on 04/18/18 from IP address 125.161.136.52. Copyright ARRS. For personal use only; all rights reserved
thus would be excluded from this study. Nod-
ules with typical radiologic findings for HCC
but that were pathologically confirmed as
non-HCCs were also excluded, making eval-
uation of the sensitivity or specificity of each
MRI finding impossible. A prospective study
with a long follow-up period might be neces-
sary to calculate specificity. Second, we ex-
cluded nine pathologically confirmed HCCs
that were not visible on all MRI sequences
because radiologic diagnosis is not possible
for nonvisible nodules in practice. Nonethe-
less, lack of visibility might be characteristic
of these nodules. Third, we defined washout
as low signal intensity on the portal or tran-
sitional phase. However, for gadoxetic acid–
enhanced MRI, it is controversial whether
low signal intensity on the transitional phase
is washout. The frequency of washout in our
study might have been exaggerated by our def-
inition. Some researchers suggest that the di-
agnosis of noninvasive HCC should be made
only when extracellular contrast agents are
used. Gadoxetic acid–enhanced MRI is the
most sensitive imaging tool for the detection
and characterization of HCC, and we used the
definition of washout from the 2014 Korean
practice guidelines for managing HCC [8].
In conclusion, HCCs smaller than 1.5 cm
in size less frequenty showed the MRI find-
ings seen typically in larger HCCs. Therefore,
small HCCs are harder to detect with certain-
ty not only because of the small size but also
because of the lower frequency of seeing typi-
cal MRI findings.
References
1. European Association for the Study of the Liver,
European Organisation for Research and Treat-
ment of Cancer. EASL-EORTC clinical practice
guidelines: management of hepatocellular carci-
noma. J Hepatol 2012; 56:908–943
2. Bruix J, Sherman M; American Association for
the Study of Liver Diseases. Management of hepa-
AJR:208, March 2017 551
MRI of Small HCC
tocellular carcinoma: an update. Hepatology
2011; 53:1020–1022
3. Kudo M, Izumi N, Kokudo N, et al. Management
of hepatocellular carcinoma in Japan: consensus-
based clinical practice guidelines proposed by the
Japan Society of Hepatology (JSH) 2010 updated
version. Dig Dis 2011; 29:339–364
4. Lee JM, Park JW, Choi BI. 2014 KLCSG-NCC
Korea practice guidelines for the management of
hepatocellular carcinoma: HCC diagnostic algo-
rithm. Dig Dis 2014; 32:764–777
5. Forner A, Vilana R, Ayuso C, et al. Diagnosis of
hepatic nodules 20 mm or smaller in cirrhosis:
prospective validation of the noninvasive diagnos-
tic criteria for hepatocellular carcinoma. Hepatol-
ogy 2008; 47:97–104
6. Kim YK, Lee YH, Kim CS, Han YM. Added diag-
nostic value of T2-weighted MR imaging to gado-
linium-enhanced three-dimensional dynamic MR
imaging for the detection of small hepatocellular
carcinomas. Eur J Radiol 2008; 67:304–310
7. Kim TK, Lee E, Jang HJ. Imaging findings of
mimickers of hepatocellular carcinoma. Clin Mol
Hepatol 2015; 21:326–343
8. Korean Liver Cancer Study Group (KLCSG), Na-
tional Cancer Center (NCC). 2014 Korean Liver
Cancer Study Group-National Cancer Center Ko-
rea practice guideline for the management of he-
patocellular carcinoma. Korean J Radiol 2015;
16:465–522
9. Mitchell DG, Bruix J, Sherman M, Sirlin CB.
LI-RADS (Liver Imaging Reporting and Data
System): summary, discussion, and consensus of
the LI-RADS Management Working Group and
future directions. Hepatology 2015; 61:1056–1065
10. Park YN. Update on precursor and early lesions of
hepatocellular carcinomas. Arch Pathol Lab Med
2011; 135:704–715
11. Veltri A, Calvo A, Tosetti I, et al. Experiences in
US-guided percutaneous radiofrequency ablation
of 44 renal tumors in 31 patients: analysis of pre-
dictors for complications and technical success.
Cardiovasc Intervent Radiol 2006; 29:811–818
12. Kim YK, Kim CS, Han YM, Park G. Detection of
small hepatocellular carcinoma: can gadoxetic
acid-enhanced magnetic resonance imaging re-
place combining gadopentetate dimeglumine-en-
hanced and superparamagnetic iron oxide-en-
hanced magnetic resonance imaging? Invest
Radiol 2010; 45:740–746
13. Lee YJ, Lee JM, Lee JS, et al. Hepatocellular car-
cinoma: diagnostic performance of multidetector
CT and MR imaging—a systematic review and
meta-analysis. Radiology 2015; 275:97–109
14. Yu MH, Kim JH, Yoon JH, et al. Small (≤1-cm)
hepatocellular carcinoma: diagnostic performance
and imaging features at gadoxetic acid-enhanced
MR imaging. Radiology 2014; 271:748–760
15. Rhee H, Kim MJ, Park YN, Choi JS, Kim KS. Ga-
doxetic acid-enhanced MRI findings of early hepa-
tocellular carcinoma as defined by new histologic
criteria. J Magn Reson Imaging 2012; 35:393–398
16. Hussain HK, Syed I, Nghiem HV, et al.
T2-weighted MR imaging in the assessment of
cirrhotic liver. Radiology 2004; 230:637–644
17. Shinmura R, Matsui O, Kobayashi S, et al. Cir-
rhotic nodules: association between MR imaging
signal intensity and intranodular blood supply.
Radiology 2005; 237:512–519
18. Honda H, Kaneko K, Maeda T, et al. Small hepa-
tocellular carcinoma on magnetic resonance im-
aging: relation of signal intensity to angiographic
and clinicopathologic findings. Invest Radiol
1997; 32:161–168
19. Ahn SS, Kim MJ, Lim JS, Hong HS, Chung YE,
Choi JY. Added value of gadoxetic acid-enhanced
hepatobiliary phase MR imaging in the diagnosis
of hepatocellular carcinoma. Radiology 2010;
255:459–466
20. Joo I, Lee JM, Lee DH, Jeon JH, Han JK, Choi BI.
Noninvasive diagnosis of hepatocellular carcinoma
on gadoxetic acid-enhanced MRI: can hypointensi-
ty on the hepatobiliary phase be used as an alterna-
tive to washout? Eur Radiol 2015; 25:2859–2868
21. Choi JY, Lee JM, Sirlin CB. CT and MR imaging
diagnosis and staging of hepatocellular carcino-
ma. Part II. Extracellular agents, hepatobiliary
agents, and ancillary imaging features. Radiology
2014; 273:30–50
22. Di Martino M, Di Miscio R, De Filippis G, et al.
Detection of small (≤2 cm) HCC in cirrhotic pa-
tients: added value of diffusion MR-imaging. Ab-
dom Imaging 2013; 38:1254–1262
23. Kojiro M. Focus on dysplastic nodules and early
hepatocellular carcinoma: an Eastern point of
view. Liver Transpl 2004; 10(2 suppl 1):S3–S8
24. Rhee H, Kim MJ, Park MS, Kim KA. Differentia-
tion of early hepatocellular carcinoma from be-
nign hepatocellular nodules on gadoxetic acid-
enhanced MRI. Br J Radiol 2012; 85:e837–e844