Hepatocellular adenomas (HCAs) are a heterogeneous group of liver tumors with variable prognosis according to subtype. Although many are indolent, certain subtypes may undergo malignant transformation or present with hemorrhage that may be life-threatening [
1]. Since the initial recognition of distinct HCA subtypes, their classification has undergone several updates [
2,
3]. The most recent update in 2017 classified HCAs into eight unique subtypes on the basis of genetic and molecular analysis, each with clinical and management implications [
4]. For example, the β-catenin exon 3 activated subtype has the highest risk of malignant transformation, and a newly recognized subtype, sonic hedgehog, has the highest risk of symptomatic bleeding. However, some HCAs do not undergo biopsy, and molecular analysis is not commonly performed clinically. Therefore, a method to identify the subtype on the basis of imaging would help risk stratify HCAs noninvasively.
Methods
Patient Sample
This single-center, retrospective, HIPAA-compliant study was approved by the institutional review board. The requirement for written informed consent was waived. We searched the institutional pathology database for consecutive adult patients (≥ 18 years) with a histologic diagnosis of HCA from January 2010 to January 2021, yielding 60 patients. Two of these patients were excluded because they underwent imaging evaluation only by CT, and two were excluded because they underwent imaging evaluation by MRI using a contrast agent other than gadoxetate disodium. These exclusions resulted in a final study sample of 56 patients (49 women, seven men; mean age, 37 ± 13 [SD] years; age range, 18–70 years). Four patients were included in a prior study that evaluated gadoxetate disodium–enhanced MRI features on the basis of the 2007 classification system [
5].
The study sample was assembled by an abdominal radiology fellow (J.R.T.) serving as the study coordinator, who reviewed the available MRI examinations, intraprocedural images from image-guided biopsies, and pathology results. Only HCAs with available tissue pathology from image-guided biopsy or surgical resection were analyzed. In patients with more than three HCAs with available tissue pathology, the three largest such HCAs were selected to reduce clustering effects. The investigator annotated on the MR images the locations of the HCAs with available tissue pathology to guide the subsequent blinded reader analysis. The investigator also recorded the size of each HCA based on its longest axial diameter on any sequence and also recorded the number of HCAs per patient regardless of whether the HCAs had available tissue pathology (classified as single, multiple [2–9 HCAs], or adenomatosis [≥ 10 adenomas]). In addition, the investigator reviewed the electronic medical record (EMR) for each patient for the presence of obesity (BMI > 30), history of oral contraceptive pill use, and history of anabolic steroid use. Finally, for HCAs that underwent surgical resection, the investigator reviewed the clinical pathology reports to record the presence of necrosis, hemorrhage, or malignant transformation to hepatocellular carcinoma (HCC).
Determination of Hepatocellular Adenoma Histologic Subtype
For purposes of this investigation, a hepatobiliary pathologist (B.V.N., with 11 years of posttraining experience) blinded to the MRI findings reviewed the tissue specimens to reclassify the HCC sub-type using the 2017 classification. For HCAs that underwent both core biopsy and surgical resection, the pathologist reviewed both specimens in a single session. All specimens had undergone H and E and immunohistochemical (IHC) staining for serum amyloid A, glutamine synthase, β-catenin, and liver fatty-acid binding protein as part of standard clinical assessment. The pathologist classified HCAs into the following subtypes on the basis of the IHC staining pattern and established algorithms [
8,
9]: hepatocyte nuclear factor-1α mutated HCA (HHCA), negative liver fatty-acid binding protein; inflammatory HCA (IHCA), positive serum amyloid A without nuclear staining of β-catenin; β-catenin exon 3 activated HCA (β-HCA), diffuse and strong positive glutamine synthase or nuclear β-catenin staining; and mixed IHCA and β-HCA (β-IHCA), diffuse and strong positive glutamine synthase or nuclear β-catenin staining plus serum amyloid A staining. For purposes of this analysis, in HCAs that could not be classified on the basis of those staining patterns, the tissue was stained with arginosuccinate synthetase 1 and classified as sonic hedgehog HCA (shHCA) if positive (defined as expression greater than that of liver [Fig. S1, available in the
online supplement]) or as unclassified HCA (UHCA) if negative, according to established algorithms [
10,
11]. Further details regarding the IHC staining for arginosuccinate synthetase 1 are described in the Supplemental Methods (available in the
online supplement).
MRI Examination Acquisition
MRI examinations were performed on a variety of 3-T (Skyra, Vida Prisma, and PrismaFIT, Siemens Healthineers; Architect, GE Healthcare) or 1.5-T (Avanto, Sonata, and Symphony, Siemens Healthineers; Signa HDxt, GE Healthcare) scanners using a single standardized liver MRI protocol. Acquisition parameters were optimized for each scanner. In all patients, unenhanced sequences included single and multishot axial and coronal 2D T2-weighted images, with and without fat suppression; axial and coronal 3D spoiled gradient-echo T1-weighted images, with and without fat suppression; and axial 2D dual-echo in- and opposed-phase T1-weighted images. After IV administration by power injector of 0.1 mL/kg body weight (0.025 mmol/kg body weight; maximum dose of 10 mL) of gadoxetate disodium (Eovist, Bayer Healthcare), dynamic 3D axial T1-weighted spoiled gradient-echo images with fat saturation were acquired in the arterial, portal venous, transitional (3 minutes), and hepatobiliary (20 minutes) phases.
MRI Analyses
Qualitative features—Two fellowship-trained abdominal radiologists with 5 and 22 years of posttraining experience (E.R.F., S.S.R., respectively) independently reviewed the MRI examinations on a workstation (Centricity, GE Healthcare). The radiologists were aware that the annotated lesions all represented HCA but were blinded to HCA subtype. Each patient was assessed in terms of the presence or absence of liver steatosis (at least 5% signal dropout on opposed-phase images according to reader-placed ROIs). Each annotated HCA was assessed for the presence or absence of the following features: intralesional steatosis (at least 5% signal dropout on opposed-phase images based on reader-placed ROIs; when present, classified as homogeneous diffuse vs heterogeneous), “atoll” sign (peripheral rim of T2 hyperintensity), hemorrhage (nonenhancing, heterogeneous, intrinsic, moderate-to-marked T1 hyperintensity, with or without susceptibility artifact and T2 hypointensity), and fluid components (nonenhancing T2 hyperintensity). HCAs were also classified on T2-weighted images as hypointense, isointense or mildly hyperintense, or moderately hyperintense relative to liver, and on precontrast as well as on arterial, portal venous, and hepatobiliary phase post-contrast T1-weighted images as hypointense, isointense, or hyperintense relative to liver. Among HCAs with hepatobiliary phase iso- or hyperintensity, the hepatobiliary phase signal intensity (SI) was further characterized as homogeneous versus heterogeneous. The two readers' independent interpretations were used only for purposes of determining interreader agreement. For further analyses, discrepancies between the two readers were resolved by consulting with a third radiologist (D.S.K.L., with 28 years of posttraining experience). For features assessed only under certain conditions (e.g., homogeneity of intralesional steatosis) that the reader did not assess during the independent evaluations, the reader evaluated the feature if the conditional feature was deemed present (e.g., intralesional steatosis) at the time that discrepancies were resolved.
Quantitative features—One of the two radiologists who performed the qualitative assessments (E.R.F.) performed an additional quantitative evaluation of HCAs on precontrast as well as arterial, portal venous, transitional, and hepatobiliary phase postcontrast T1-weighted images. This evaluation was performed in a separate session, 6 weeks after completion of the qualitative readings. The investigator placed ROIs with a diameter of at least 1 cm on HCAs on a single axial slice, avoiding hemorrhage or fluid components. At least three ROIs were placed on each HCA (depending on lesion size), and the mean SI was determined (SI
HCA). The investigator also placed ROIs with a diameter of at least 1 cm on adjacent liver parenchyma, avoiding vasculature or bile ducts. Three ROIs were placed on the liver, and the mean SI was determined (SI
liver). The SI ratio was calculated for precontrast and all postcontrast phases as the ratio between SI
HCA and SI
liver. The liver-to-lesion contrast enhancement ratio (LLCER) was also determined for all postcontrast phases using the following formula [
12,
13]:
where “-Post” indicates mean postcontrast SI ratio and “-Pre” indicates mean precontrast SI ratio.
Post Hoc Assessment of Hepatocellular Adenomas With Malignant Degeneration to Hepatocellular Carcinoma
The study coordinator (J.R.T.) performed a post hoc assessment of HCAs with malignant degeneration to HCC on pathologic evaluation. For these HCAs, the investigator recorded HCA size, the size of the HCC component on pathologic assessment, the degree of differentiation of HCC on pathologic assessment, and the presence versus absence of areas of hepatobiliary phase isoor hyperintensity.
Statistical Analysis
Qualitative features were expressed using counts with percentages, and quantitative features were expressed using mean and SD. Interreader agreement of qualitative features was assessed using weighted Cohen kappa coefficients for ordinal measures and Cohen kappa coefficients for binary measures. Agreement was classified as follows: poor, less than 0.20; fair, 0.20–0.39; moderate, 0.40–0.59; substantial, 0.60–0.79; almost perfect, 0.80–0.99; or perfect, 1.00.
The study coordinator (J.R.T.) performed an exploratory analysis of the clinical characteristics or qualitative or quantitative MRI features that subjectively distinguished individual HCA subtypes. Identified features were then compared between the potentially associated HCA subtype and all other HCA subtypes combined. These comparisons used Fisher exact test for qualitative features and
t test for quantitative features. HCAs showing hepatobiliary phase iso- or hyperintensity or showing a positive (i.e., > 0) LLCER were also further summarized. The investigator then generated a proposed stepwise diagnostic algorithm for determining HCA subtype according to gadoxetate disodium–enhanced MRI features based on findings from the present analysis and relevant earlier studies [
14–
16]. The algorithm's sensitivity, specificity, and accuracy for determining HCA subtypes in the present cohort were calculated.
For purposes of the exploratory analysis and the proposed diagnostic algorithm, HCAs with any β-catenin exon 3 mutations were combined (i.e., β-HCA and β-IHCA). All comparisons were two sided, and p values less than .05 were considered statistically significant. Statistical analysis was performed using GraphPad Prism version 9.1.2 (GraphPad Software) and Stata 16.1 (Stata).
Results
Patients and Tissue Pathology
The 56 patients had a total of 65 HCAs with available tissue pathology and that were included in the analysis. Of the 65 HCAs, tissue was obtained only by core biopsy in 56 and by both core biopsy and surgical resection in 33. In all HCAs that underwent both core biopsy and surgical resection, the subtype was the same according to pathologic assessment of both the core biopsy and the surgical specimens. A total of 48 patients had one HCA, seven patients had two HCAs, and one patient had three HCAs. In patients with multiple HCAs, all HCAs were of the same subtype.
According to pathologic assessment, the 65 HCAs included 16 (25%) HHCAs, 31 (48%) IHCAs, six (9%) β-HCAs, four (6%) β-IHCAs, five (8%) shHCAs, and three (5%) UHCAs. Of the 33 HCAs that underwent surgical resection, the pathology report indicated the presence of necrosis in four (one IHCA, two β-HCA, one shHCA), hemorrhage in 16 (eight IHCA, one HHCA, one shHCA, three β-HCA, and three β-IHCA), and malignant transformation in three (one HHCA, two β-HCA).
Figure 1 shows the flow of patient selection and the distribution of HCA subtypes in the final sample.
Table 1 summarizes patient and HCA characteristics stratified by HCA subtype.
Exploratory Assessment of Associations of Hepatocellular Adenoma Subtypes With Clinical and Noncontrast MRI Features
Table 2 summarizes the qualitative MRI features stratified by HCA subtype. Interreader agreement was substantial to almost perfect for all qualitative features (0.60–0.92), except for moderate agreement for the atoll sign (0.53).
HHCAs showed homogeneous diffuse intralesional steatosis in 94% (15/16), whereas all other HCAs combined showed homogeneous diffuse intralesional steatosis in 0% (0/49) (p < .001). In patients with HHCA, liver steatosis was present in 8% (1/13), whereas patients with all other HCA subtypes combined had liver steatosis in 45% (24/53) (p = .009).
IHCAs showed the atoll sign in 58% (18/31), whereas all other HCAs combined showed the atoll sign in 12% (4/34) (p < .001). IHCAs showed moderate T2 hyperintensity in 52% (16/31), whereas all other HCAs combined showed moderate T2 hyperintensity in 12% (4/34) (p < .001).
Subtypes β-HCA and β-IHCA occurred in men in 63% (5/8), whereas all other HCAs combined occurred in men in 4% (2/48) (p < .001). Subtypes β-HCA and β-IHCA had a mean size of 10.1 ± 6.8 cm, whereas all other HCAs combined had a mean size of 5.1 ± 2.9 cm (p = .03). Subtypes β-HCA and β-IHCA showed hemorrhage in 40% (4/10), whereas all other HCAs combined showed hemorrhage in 11% (6/55) (p = .04). Subtypes β-HCA and β-IHCA showed fluid components in 60% (6/10), whereas all other HCAs combined showed fluid components in 5% (3/55) (p < .001).
No distinguishing clinical or qualitative MRI features were identified for shHCAs or UHCAs.
Exploratory Assessment of Associations of Hepatocellular Adenoma Subtypes With Multiphase Postcontrast MRI Features
Table 3 summarizes mean SI ratios and mean LLCERs, stratified by HCA subtype. Figures S2 and S3 (available in the
online supplement) depict the distribution of individual values for SI ratio and LLCER, respectively, stratified by phase and HCA subtype, and also plot the mean values for SI ratio and LLCER across phases.
On precontrast T1-weighted fat-saturated images, hypointensity was observed in 100% (16/16) of HHCAs compared with 10% (5/49) of all other HCAs combined (p < .001). The precontrast SI ratio was significantly lower for HHCAs (0.69 ± 0.14) than for all other HCAs combined (1.05 ± 0.22) (p < .001).
On arterial phase images, hyperintensity was observed in 50% (8/16) of HHCAs compared with 92% (45/49) of all other HCAs combined (p < .001). The arterial phase SI ratio was significantly lower for HHCAs (1.24 ± 0.28) than for all other HCAs combined (1.48 ± 0.38) (p = .02). However, the arterial phase LLCER was significantly higher for HHCAs (0.90 ± 0.70) than for all other HCAs combined (0.42 ± 0.29) (p < .001).
On portal venous phase images, hypointensity was observed in 94% (15/16) of HHCAs compared with 16% (8/49) of all other HCAs combined (p < .001). The portal venous phase SI ratio was significantly lower for HHCAs (0.81 ± 0.15) than for all other HCAs combined (1.05 ± 0.15) (p < .001). However, the portal venous phase LLCER was significantly higher for HHCAs (0.22 ± 0.37) than for all other HCAs combined (0.02 ± 0.15) (p = .003).
On transitional phase images, the SI ratio was significantly lower for HHCAs (0.68 ± 0.14) than for all other HCAs combined (0.95 ± 0.19) (p < .001). The LLCER did not distinguish any HCA subtype.
On hepatobiliary phase images, iso- or hyperintensity was observed in 80% (8/10) of β-HCAs and β-IHCAs compared with 5% (3/55) of all other HCAs combined (p < .001). The hepatobiliary phase SI ratio was significantly higher for β-HCAs and β-IHCAs (1.04 ± 0.21) than for all other HCAs combined (0.64 ± 0.16) (p < .001), as well as significantly lower for HHCAs (0.46 ± 0.12) than for all other HCAs combined (0.78 ± 0.19) (p < .001). The hepatobiliary phase LLCER was significantly higher for β-HCAs and β-IHCAs (0.04 ± 0.29) than for all other HCAs combined (–0.30 ± 0.14) (p < .001).
The three HCAs other than β-HCA and β-IHCA that showed hepatobiliary phase iso- or hyperintensity were IHCAs. Two of these three IHCAs were in patients with steatosis, and both of these IHCAs showed hyperintensity on precontrast T1-weighted MR images. The hepatobiliary phase LLCER was positive (i.e., greater than zero) in nine HCAs (eight β-HCAs or β-IHCAs and the one IHCA in a patient without steatosis), and negative in the remaining 56 HCAs.
Characteristic findings on gadoxetate disodium–enhanced MRI are shown in an example of HHCA in
Figure 2, IHCA in Figure S4, β-HCA in
Figure 3, shHCA in Figure S5, and UHCA in Figure S6 (Figs. S4–S6 are available in the
online supplement). The one IHCA with LLCER greater than 0 is shown in Figure S7 (available in the
online supplement).
Proposed Diagnostic Algorithm
Figure 4 presents the proposed stepwise diagnostic algorithm for determining HCA subtype according to gadoxetate di-sodium–enhanced MRI features. First, if the HCA shows homogeneous diffuse intralesional steatosis, then HHCA is suspected (sensitivity, 94% [15/16]; specificity, 100% [49/49]; accuracy, 98% [64/65]). If HHCA is not suspected, then if the HCA shows either the atoll sign or moderate T2 hyperintensity, IHCA is suspected (sensitivity, 74% [23/31]; specificity, 91% [31/34]; accuracy, 83% [54/65]). If neither HHCA or IHCA is suspected, then if the HCA shows hepatobiliary phase hyperintensity, β-HCA or β-IHCA is suspected (sensitivity, 80% [8/10]; specificity, 98% [54/55]; accuracy, 95% [62/65]). Hepatobiliary phase hypointensity may in general be assessed qualitatively, though LLCER is used in patients with steatosis. If HHCA, IHCA, or β-HCA or β-IHCA are not suspected, then the HCA is considered indeterminate by MRI, and biopsy should be considered. According to the application of the algorithm to the present sample, indeterminate HCAs would be IHCA in 47% (7/15), shHCA in 33% (5/15), β-HCA or β-IHCA in 7% (1/15), and UHCA in 13% (2/15).
Post Hoc Assessment of Hepatocellular Adenoma With Malignant Transformation to Hepatocellular Carcinoma
The three HCAs with malignant transformation to HCC had sizes of 8.7, 16.0, and 16.1 cm. In all three lesions, the HCC component was well differentiated. The HCC component measured at least 1 cm in two lesions (both β-HCA; HCC component measuring 3.5 and 10.3 cm) and less than 1 cm in one lesion (HHCA; HCC component measuring 0.8 cm). Areas of hepatobiliary phase iso- or hyper-intensity were observed in both β-HCAs with malignant transformation but not in the HHCA with malignant transformation.
Discussion
In this study, we sought to identify distinguishing gadoxetate disodium–enhanced MRI features of HCA subtypes on the basis of the most recent HCA genotypic classification system described by Nault et al. [
4]. To our knowledge, the study sample comprises the largest number of HCAs with β-catenin exon 3 mutations to be characterized by gadoxetate disodium–enhanced MRI. Findings more strongly associated with HHCAs than with other sub-types included intralesional steatosis, absence of liver steatosis, hypointensity in the portal venous phase, and low SI ratio across precontrast and all postcontrast phases. Findings more strongly associated with IHCAs than with other subtypes included the atoll sign and moderate T2 hyperintensity. Findings more strongly associated with β-HCA and β-IHCA compared with the other subtypes included male sex, large size, hemorrhage, fluid components, and hepatobiliary phase iso- to hyperintensity. According to the present results and those of earlier studies, we proposed a diagnostic algorithm for identifying the three most common subtypes (IHCA, HHCA, and β-HCA or β-IHCA). The shHCAs and UHCAs were the two least common subtypes in the study sample and did not show distinguishing MRI features; nonetheless, all such HCAs showed hepatobiliary phase hypointensity.
The proposed diagnostic algorithm first identifies HHCAs, followed by IHCAs. These are the two most common subtypes, each accounting for 30–35% of all HCAs [
4]. The European Association for the Study of the Liver guidelines indicate that both subtypes can be accurately diagnosed by MRI [
4]. According to the algorithm, HHCAs are identified by the presence of homogeneous diffuse intralesional steatosis, and IHCAs are identified by the presence of the atoll sign or moderate T2 hyperintensity [
14–
16]. The remaining HCAs are evaluated in terms of presence or absence of hepatobiliary phase iso- or hyperintensity. As a result of this stepwise approach, HCAs are characterized as IHCAs if showing atoll sign or moderate T2 hyperintensity, even if also showing hepatobiliary phase iso- or hyperintensity. The identification of IHCAs on the basis of atoll sign or moderate T2 hyperintensity before evaluation of hepatobiliary phase characteristics is intended to improve the specificity of hepatobiliary phase iso- or hyperintensity for β-HCA or β-IHCA. External validation of the algorithm across institutions is required.
HCAs were evaluated in terms of both HCA-to-liver SI ratio and LLCER. Of these two measures, LLCER better reflects a lesion's true enhancement pattern relative to liver because it accounts for the impact of the liver's precontrast SI, including the presence of steatosis [
17]. For example, 94% of HHCAs showed hypointensity on portal venous phase images, and portal venous phase SI ratio was significantly lower for HHCA than for other subtypes. However, on portal venous phase images, LLCER was significantly higher for HHCA than for other subtypes, suggesting signal suppression of fat-containing lesions as a possible explanation for the apparent portal venous washout of HHCAs. Similarly, of the three IHCAs with hepatobiliary phase iso- or hyperintensity, only one had a positive LLCER. This observation suggests that this IHCA was the only one with true gadoxetate disodium retention, and that the apparent hepatobiliary phase iso- or hyperintensity of the other two IHCAs, which both had negative LLCER, may be attributable to the presence of hepatic steatosis. LLCER is more time-consuming to measure than SI ratio given the larger number of ROIs and thus may be less practical for routine clinical implementation. We believe that simple qualitative assessment or SI ratios are sufficient for assessing most HCAs, and indeed, the proposed diagnostic algorithm does not formally incorporate LLCER. However, LLCER may be useful in select cases, for example, when the algorithm suggests that an HCA with hepatobiliary phase iso- or hyperintensity is IHCA. In this scenario, LLCER could potentially be used to assess whether the hepatobiliary phase iso- or hyperintensity may be attributable to liver steatosis [
17].
Gadoxetate disodium is a hepatocyte-specific MRI contrast agent that may help determine HCA subtype. It is predominantly transported into hepatocytes by organic anion-transporting polypeptides (OATP) 1B1/1B3 transporters at the basolateral membrane. Additionally, multidrug resistance-associated protein 3 (MRP3), also expressed at the basolateral membrane, allows complementary sinusoidal backflux, and MRP2, expressed at the canalicular membrane, excretes the contrast material into the bile ducts [
6,
18]. Among the three proteins, hepatobiliary phase SI in HCAs appears to be mostly determined by OATP 1B1/1B3 expression; preserved or increased expression is associated with hepatobiliary phase iso- or hyperintensity, whereas decreased or absent expression is associated with hypointensity [
6,
18–
20]. Variation among subtypes in genotypic mutations may account for variable expression of OATP 1B1/1B3, in turn accounting for variation in hepatobiliary phase findings. For example, in HHCAs, biallelic inactivation of the transcription factor hepatocyte nuclear factor-1α leads to down-regulation of liver fatty-acid binding protein and OATP [
21,
22], which may lead to intralesional fat accumulation and marked hepatobiliary phase hypointensity, respectively. On the other hand, in β-HCA, increased Wnt/β-catenin signaling is associated with increased OATP 1B1/1B3 expression, possibly leading to hepatobiliary phase iso- or hyperintensity [
6,
18–
20,
23–
25].
Prior studies have also recognized potential hepatobiliary phase iso- or hyperintensity in IHCAs, mimicking FNH or β-HCA [
26,
27], and a meta-analysis from 2022 [
28] reported the pooled rate of hepatobiliary phase iso- or hyperintensity of IHCAs to be 14% (95% CI, 2–31%). However, steatosis that is common in patients with IHCA may cause the HCA to show apparent increased hepatobiliary phase SI despite no true contrast material retention. For example, in a study of 17 IHCAs with hepatobiliary phase iso- to hyperintensity, all had hepatic steatosis, but none had a positive LLCER [
17]. In addition, at immunohistochemistry analysis, Sciarra et al. [
19] reported that 87% of IHCAs had deceased expression of OATP 1B1/1B3 compared with liver, 13% had similar expression, and none had overexpression; on the other hand, all β-HCAs had overexpression. Thus, some prior studies may have overestimated the frequency of true gadoxetate disodium retention among IHCAs.
A major complication of HCAs is malignant transformation to HCC, particularly for large HCAs (i.e., greater than 5 cm). The risk is highest for HCAs with β-catenin exon 3 mutations, although malignant transformation has been reported in all subtypes [
4,
29]. Among cases of malignant degeneration of HCA, 58% present as a small focus (i.e., less than 1 cm) [
29]. In addition, in a propensity-matched sample, HCC developing in a HCA had generally favorable prognosis compared with spontaneous HCCs. In this study, malignant transformation occurred in three HCAs: two β-HCA (both with large foci, measuring > 1 cm) and one HHCA (with a small focus < 1 cm). HHCAs have a low but well-documented risk of malignant transformation. For example, Nault et al. [
4] identified HCC in 5% (7/133) of HHCAs, and Putra et al. [
30] found that HHCAs more commonly showed malignant transformation in women (whereas β-HCA more commonly showed malignant transformation in men). Although the exact mechanism of malignant transformation is unknown, it is not believed to involve the β-catenin signaling pathway [
30]. Indeed, in the current study, the MR images of the HHCA with malignant transformation did not show any areas of hepatobiliary phase iso- or hyper-intensity (a finding associated with β-catenin mutations).
This study had limitations. First, the study used a retrospective single-institution design, and some HCA subtypes had small sample sizes. The number of IHCAs with hepatobiliary phase isoor hyperintensity was also small. Further studies with a prospective multiinstitutional design are needed to help validate the results and the proposed diagnostic algorithm. In addition, the diagnostic algorithm did not account for other liver lesions that may exhibit hepatobiliary phase iso- or hyperintensity, such as FNH, although prior studies have addressed the differentiation of HCA and FNH using features on gadoxetate disodium–enhanced MRI [
31,
32]. Also, exon sequencing was not used to define HCA subtypes. However, immunohistochemistry has been shown to be robust for identifying HCA subtypes and is widely used clinically [
8,
9]. Moreover, three HCAs remained unclassified after immunohistochemistry analysis, and additional testing was not performed to determine whether these represented β-catenin exon 7/8 HCAs or were truly unclassified HCAs. Finally, the statistical comparisons did not account for the small number of patients with multiple HCAs.
In conclusion, we identified distinguishing characteristics of HCA subtypes on gadoxetate disodium–enhanced MRI according to the 2017 genotypic classification of HCA subtypes. Common subtypes (IHCA, HHCA, and β-HCA or β-IHCA) could be identified by a combination of MRI findings, including intralesional steatosis, the atoll sign, moderate T2 hyperintensity, and hepatobiliary phase iso- or hyperintensity. A newly described subtype, shHCA, was hypointense on the hepatobiliary phase but otherwise did not show characteristic MRI features. LLCER may be useful to help assess whether apparent hepatobiliary phase iso- or hyperintensity in a suspected IHCA may be attributed to hepatic steatosis. A diagnostic algorithm was proposed for determining HCA sub-type according to gadoxetate disodium–enhanced MRI findings and could identify the most common subtypes with high accuracy, including HCAs with β-catenin exon 3 mutations. Further validation of the algorithm is required.