Hepatocellular adenomas: Magnetic resonance imaging features as a function of molecular pathological classification

Authors

  • Hervé Laumonier,

    Corresponding author
    1. Department of Radiology, CHU de Bordeaux, Hŏpital Saint-André, Bordeaux, France
    • Department of Radiology, CHU de Bordeaux, Hŏpital Saint-André, Bordeaux, F-33076, France
    Search for more papers by this author
  • Paulette Bioulac-Sage,

    1. Department of Pathology, CHU de Bordeaux, Hŏpital Saint-André, Bordeaux, France
    2. Department of Surgery, CHU de Bordeaux, Hŏpital Saint-André, Bordeaux, France
    Search for more papers by this author
  • Christophe Laurent,

    1. Department of Hepatology, CHU de Bordeaux, Hŏpital Saint-André, Bordeaux, France
    2. INSERM, U889, and Université Victor Segalen Bordeaux 2, Bordeaux, France
    Search for more papers by this author
  • Jessica Zucman-Rossi,

    1. Laboratory for Molecular and Functional Imaging, UMR5231 CNRS and Université Victor Segalen Bordeaux 2, Bordeaux, France
    2. INSERM, U674, Génomique fonctionnelle des tumeurs solides, Paris, France
    Search for more papers by this author
  • Charles Balabaud,

    1. Department of Surgery, CHU de Bordeaux, Hŏpital Saint-André, Bordeaux, France
    2. Department of Hepatology, CHU de Bordeaux, Hŏpital Saint-André, Bordeaux, France
    3. INSERM, U889, and Université Victor Segalen Bordeaux 2, Bordeaux, France
    4. Laboratory for Molecular and Functional Imaging, UMR5231 CNRS and Université Victor Segalen Bordeaux 2, Bordeaux, France
    5. INSERM, U674, Génomique fonctionnelle des tumeurs solides, Paris, France
    6. Université Paris Diderot—Paris 7, Institut Universitaire d'Hématologie, Paris, France
    Search for more papers by this author
  • Hervé Trillaud

    1. Department of Radiology, CHU de Bordeaux, Hŏpital Saint-André, Bordeaux, France
    2. Department of Pathology, CHU de Bordeaux, Hŏpital Saint-André, Bordeaux, France
    3. Department of Surgery, CHU de Bordeaux, Hŏpital Saint-André, Bordeaux, France
    4. Department of Hepatology, CHU de Bordeaux, Hŏpital Saint-André, Bordeaux, France
    5. INSERM, U889, and Université Victor Segalen Bordeaux 2, Bordeaux, France
    Search for more papers by this author

  • Potential conflict of interest: Nothing to report.

Abstract

Hepatocellular adenomas (HCAs) are a group of benign tumors forming three molecular pathological subgroups: (1) hepatocyte nuclear factor 1α (HNF-1α)–inactivated, (2) β-catenin–activated, and (3) inflammatory. Some HCAs present both β-catenin activation and inflammation. We analyzed magnetic resonance imaging (MRI) data for correlations between features on imaging and pathological classification of HCAs. We included 50 cases for which pathology specimens were classified into three groups based on immunohistochemical staining. Two characteristic MRI profiles were identified corresponding to HNF-1α–inactivated and inflammatory HCAs. Fifteen HCAs were HNF-1α–inactivated. The corresponding lesions showed (1) diffuse signal dropout on T1-weighted chemical shift sequence due to steatosis, (2) isosignal or slight hypersignal on T2-weighted (T2W) images, and (3) moderate enhancement in the arterial phase, with no persistent enhancement in the portal venous and delayed phases. For the diagnosis of HNF-1α–inactivated HCA, the positive predictive value of homogeneous signal dropout on chemical shift images was 100%, the negative predictive value was 94.7%, the sensitivity was 86.7%, and the specificity was 100%. Twenty-three HCAs were inflammatory and showed (1) an absence or only focal signal dropout on chemical shift sequence; (2) marked hypersignal on T2W sequences, with a stronger signal in the outer part of the lesions, correlating with sinusoidal dilatation areas; and (3) strong arterial enhancement, with persistent enhancement in the portal venous and delayed phases. Marked hypersignal on T2W sequences associated with delayed persistent enhancement had a positive predictive value of 88.5%, a negative predictive value of 84%, a sensitivity of 85.2%, and a specificity of 87.5% for the diagnosis of inflammatory HCA. Conclusion: HNF-1α–mutated HCAs and inflammatory HCAs were associated with specific MRI patterns related to diffuse fat repartition and sinusoidal dilatation, respectively. (HEPATOLOGY 2008.)

Hepatocellular adenoma (HCA) is a rare type of benign monoclonal liver neoplasm, occurring mostly in young women using oral contraceptives. Several molecular features associated with HCA have recently been described (for review, see Rebouissou et al.1).

Biallelic-inactivating mutations of the TCF1 gene inactivating hepatocyte nuclear factor 1α (HNF-1α) have been identified in 35% to 50% of HCAs.2–4 Furthermore, HCAs harboring HNF-1α mutations have been shown to display repressed gluconeogenesis, together with activation of glycolysis, citrate shuttle, and fatty acid synthesis, resulting in high predicted rates of lipogenesis.5 Liver fatty acid binding protein (L-FABP), which encodes L-FABP 1, has also been shown to be silenced in these tumors, suggesting that impaired fatty acid trafficking may also contribute to the fatty phenotype of HNF-1α–inactivated HCA.

A β-catenin mutation leading to activation of the Wnt/β-catenin pathway has also been identified in 15% to 18% of HCA cases.2–4, 6 Activating mutations of β-catenin are also found in 20% to 40% of hepatocellular carcinomas (HCCs), suggesting that β-catenin is the most frequently activated oncogene in HCC7 (for review, see Laurent-Puig and Zucman-Rossi8). We have observed no case of HCA with both β-catenin mutations and biallelic inactivation of HNF-1α, suggesting that these two tumorigenic pathways are mutually exclusive. Our series of HCA cases and other studies have shown that HCAs displaying β-catenin activation were at greater risk of malignant transformation than other subtypes of HCA.3, 4, 9

A third group of HCA cases characterized by inflammatory infiltrates, sinusoidal dilation (up to telangiectasia, the term inflammatory/telangiectatic HCA is often used), dystrophic vessels, and ductular reaction has also been identified and represents 25% to 35% of all cases.3, 4 These tumors, which are frequently associated with obesity and alcohol consumption, display strong expression of the acute-phase inflammatory response, with high levels of both messenger RNA and protein for serum amyloid A (SAA) and C-reactive protein (CRP).4 About one in six cases of inflammatory HCA also display β-catenin mutation.

Based on this classification, we recently identified a panel of four antibodies for the classification of HCA with high sensitivity and specificity4: HNF-1α–inactivated HCAs do not express L-FABP; β-catenin-activated HCAs display glutamine synthetase (GS) overexpression and have β-catenin in both the cytoplasm and the nucleus; and inflammatory HCAs are characterized by the overexpression of SAA and CRP by tumoral hepatocytes. Less than 10% of HCAs express none of these markers. In our experience, less than 5% of HCAs are not suitable for genotype/phenotype analysis due to massive hemorrhage and/or necrosis.

The highly variable appearance of HCA on computed tomography (CT), magentic resonance imaging (MRI), and contrast-enhanced ultrasound scans is well documented and probably results from differences in histological features.10–17 MRI is considered the most comprehensive and noninvasive imaging work-up for HCA diagnosis.11, 12, 15, 16 MRI findings include fatty, necrotic, and hemorrhagic components, but a homogeneous hypervascular appearance may also be observed. These characteristics overlap with the appearance on MRI of other hepatic tumors, such as HCC and atypical focal nodular hyperplasia (FNH). Radiological diagnosis of HCA is often a diagnosis of elimination, and a tissue biopsy is still mandatory in the absence of firm criteria for another hepatic tumor.

Each HCA subtype is potentially associated with different risk factors, at least for HCC, and may have various patterns of progression. It would therefore be interesting to develop a method for identifying HCA subtypes without the need for a liver biopsy or resection, to facilitate large clinical studies. We therefore performed a retrospective analysis of 50 HCA cases with the aim of identifying MRI features specifically associated with each HCA subtype.

Abbreviations

CRP, C-reactive protein; FNH, focal nodular hyperplasia; GS, glutamine synthetase; HCA, hepatocellular adenoma; HCC, hepatocellular carcinoma; HNF-1α, hepatocyte nuclear factor 1α; L-FABP, liver fatty acid binding protein; MRI, magnetic resonance imaging; SAA, serum amyloid A; T1W, T1-weighted; T2W, T2-weighted.

Patients and Methods

Among the 124 HCA patients undergoing surgery at our institution or affiliated hospital (from 1984 to September 2007), we identified 48 cases in which liver MRI scans (performed between 1996 and September 2007) and suitable pathological material were available. The other 76 cases were excluded from this study due to a lack of good MRI data (lack of adequate sequences; n = 70) or nodules that were completely hemorrhagic and/or necrotic (n = 6). We were also following another 20 HCA patients that had not undergone resection (old and recent cases); liver biopsy had been performed and appropriate MRI data were available for two of these cases. We therefore included a total of 50 patients in the study, 30 of whom have been reported previously.4 Medical, surgical, pathological, and radiological data were collected and reviewed.

Histological Analysis.

All tissue specimens (48 surgical specimens, 2 biopsy specimens) were processed as follows: hematoxylin and eosin, Masson's trichrome, reticulin and Perls staining, together with CK 7 and immunostaining for patho/molecular pathological classification that is L-FABP, β-catenin, GS, SAA, and CRP, as previously described.4 We used the following staining data to classify the nodules (Table 1).

Table 1. Main Clinical Radiological and Pathological Data
Patient No.Age/SexClinical Biological DataNumber of NodulesSize of Nodule(s) Analyzed (mm)Main Pathological Data (Tumors and Nontumoral Liver)
  1. Abbreviations: LFT, liver function tests; NTL, nontumoral liver.

127/FBy chance289Steatotic HCA L-FABP negative; NTL normal
241/FPain>1028Steatotic HCA L-FABP negative; NTL normal
332/FPain380Steatotic HCA L-FABP negative; NTL normal
429/FBy chance180Steatotic HCA L-FABP negative; NTL normal
543/FBy chance150Steatotic HCA L-FABP negative; (+ major postbiopsy hemorrhage); NTL normal
639/FAbnormal LFT170Steatotic HCA L-FABP negative; NTL normal
754/FAbnormal LFT>1023Steatotic HCA L-FABP negative; NTL normal
828/FHemorrhage>1075Steatotic HCA L-FABP negative; (+ hemorrhagic necrosis areas); NTL normal
949/FPain>1090Steatotic HCA L-FABP negative; NTL normal
1032/FPain550Steatotic HCA L-FABP negative; NTL normal
1144/FBy chance>1063Steatotic HCA L-FABP negative; NTL normal
1240/FBy chance135Steatotic HCA L-FABP negative; NTL normal
1334/FPain435Steatotic HCA L-FABP negative; NTL normal
1447/FBy chance143Steatotic HCA L-FABP negative; (+ hemorragic areas); NTL normal
1543/FPain837Steatotic HCA L-FABP negative; (+ postbiopsy hemorrhage); NTL normal
1642/FBy chance370Inflammatory HCA, SAA +; NTL normal
1738/FPain155Inflammatory HCA, SAA +; NTL normal
1831/FBy chance191Inflammatory HCA, SAA +; NTL stealosis <1/3
1936/MBy chance157Inflammatory HCA, SAA and β-catenin/GS +; NTL normal
2038/FBy chance4100Inflammatory HCA, SAA + (in part hemmorrhagic and necrotic and steatosic < 1/3); NTL normal
2145/FBy chance230Inflammatory HCA, SAA + (sinusoids mildly dilated); NTL normal
2245/FBy chance>1045 (biopsy)Liver biopsy; inflammatory HCA, SAA +; NTL steatosis <1/3
2349/FPain570Inflammatory HCA, SAA +; NTL normal
2437/FBy chance>1060Major sinusoidal dilatation, weakly SAA +; NTL normal
2534/FBy chance560,6Inflammatory HCA, SAA + (hemorrhagic areas and steatosis < 1/3); NTL steatosis > 2/3
2641/FPain628Inflammatory HCA, SAA +; NTL abnormal
2740/FAbnormal LFT170Inflammatory HCA, SAA +; NTL normal
2842/FFollow-up for lymphoma361 (biopsy)Liver biopsy: inflammatory HCA, SAA +; NTL steatosis <1/3
2954/FBy chance180Inflammatory HCA, SAA +; NTL not available
3041/FBy chance670Inflammatory HCA, SAA +; NTL steatosis <1/3
3134/FBy chance2120Inflammatory HCA, SAA +; NTL normal
3251/FBy chance690Inflammatory HCA, SAA + (central hemorrhagic area); NTL steatosis > 2/3
3353/MBy chance165Inflammatory HCA, SAA + (Sinusoids mildly dilated); NTL steatosis > 2/3
3455/FBy chance125Inflammatory HCA, SAA + (steatosis < 1/3); NTL normal
3526/FAbnormal LFT1121Inflammatory HCA, SAA +; NTL normal
3642/FBy chance290Inflammatory HCA, SAA + (stealosis < 1/3); NTL not available
3743/FBy chance460Inflammatory HCA, SAA +; NTL steatosis 1/3–2/3
3839/FBy chance170Inflammatory HCA, SAA +; NTL not available
3947/FAbnormal LFT5107,45Inflammatory HCA, SAA + and β-catenin GS +; NTL normal
4035/FPain520Inflammatory HCA, SAA + and β-catenin/GS +; NTL normal
4145/FPain>1060Inflammatory HCA, SAA + and β-catenin/GS +; NTL normal
4223/FBy chance180HCA with no specificity except for the presence of areas with numerous veins, abnormal GS by IHC (β-catenin mutation); NTL normal
4321/FBy chance190HCA with no specificity except for the presence of areas with numerous veins; abnormal GS by IHC (β-catenin mutation); NTL normal
4444/FHemorrhage5109HCA: no inflammation/sinusoidal dilatation (architectural changes due to necrosis, hemorrhage, fibrosis); NTL abnormal
4547/FBy chance1102HCA: no inflammation, mild sinusoidal dilatation (architectural changes due to necrosis, hemorrhage, fibrosis); NTL stealosis < 1/3
4628/FHemorrhage150HCA: no inflammation/sinusoidal dilatation (architectural changes due to necrosis, hemorrhage, fibrosis); NTL normal
4724/FBy chance130HCA: no inflammation/sinusoidal dilatation (steatotic rim); NTL steatosis < 1/3
4823/FAbnormal LFT180HCA: no inflammation/sinusoidal dilatation; NTL normal
4958/FHemorrhage230HCA: no inflammation/sinusoidal dilatation (areas of necrosis); NTL steatosis 1/3–2/3
5044/FBy chance348HCA: no inflammation, no telangieclasia (post biopsy subcapsular hematoma, minor necrotic/hemorrhagic changes); NTL normal (potal embolization)

Fifteen HCAs were L-FABP–negative (patients 1 to 15), corresponding to the HNF-1α–inactivated subtype and strongly associated with marked steatosis.

Twenty-seven HCAs were inflammatory (patients 16 to 41, with two nodules analyzed for patient 26). SAA was either homogeneously or heterogeneously overexpressed, often with more intense staining at the periphery, contrasting with the surrounding liver. The intensity of inflammation and sinusoidal dilatation varied among cases. Most cases displayed a typical pattern, with an obvious telangiectatic phenotype.3, 4 In one case with intense sinusoidal dilatation and minimal inflammation, SAA immunostaining was extremely faint. All these inflammatory HCAs also overexpressed CRP. This group of inflammatory HCA cases included three cases (patients 39 to 41) with an additional β-catenin activation.

Two HCAs displayed β-catenin activation without inflammation (patients 42 and 43). They displayed diffuse GS overexpression, differing in intensity, and aberrant cytoplasmic and nuclear β-catenin expression.4 There was no pathological, biological, or clinical evidence to support the classification of these HCAs as HCC or borderline lesions.3

Seven HCAs remained unclassified (patients 44 to 50), with normal L-FABP immunostaining and no staining for SAA, GS, or β-catenin. No specific pathological feature (no significant steatosis, inflammation, or sinusoidal dilatation) or specific immunostaining characteristics were found.

Thirty patients presented multiple HCAs on imaging; pathological examination of several nodules was possible in 25 of these patients. If fewer than three nodules were found in a single patient, all were subjected to immunohistochemical staining. If more than three nodules were found or the patient displayed adenomatosis, then at least three nodules were studied.

The pathological classification was validated by a molecular analysis searching for HNF-1α or β-catenin mutations in 11 and 5 of the cases, respectively. No HNF-1α or β-catenin mutation was detected by molecular analysis in any of the unclassified cases.

MRI.

All examinations were performed with a 1.5-T closed magnet system. Thirty-two patients underwent preoperative MRI examination at Bordeaux University Hospital (Sonata, Siemens Medical System, Erlangen, Germany; Achieva, Philips Medical Systems, Best, The Netherlands) with a dedicated abdominal phased array coil for signal reception. The following sequences were acquired and analyzed: axial in-phase and out-of-phase chemical shift GRE T1-weighted (T1W) images (repetition time/echo time, 208/2.3 and 4.6 msec; flip angle, 80°; field of view, 430 mm; matrix, 292 × 178; number of sections, 30; section thickness, 5.4 mm; two signals acquired); a respiratory-triggered, fat-suppressed, T2-weighted (T2W) fast spin-echo pulse sequence (repetition time/echo time, 1,287/70 msec; flip angle, 90°; field of view, 450 mm; matrix, 308 × 156; number of sections, 30; section thickness, 5 mm; one signal acquired); a T2W fast spin-echo pulse sequence (repetition time/echo time, 531/60 msec; flip angle, 90°; field of view, 395 mm; matrix, 256 × 136; number of sections, 25; section thickness, 6 mm; one signal acquired); and fat-suppressed dynamic gadolinium-enhanced T1W gradient echo sequences during the arterial phase, late arterial phase, portal venous phase, and delayed phase, with manual administration of gadolinium-based contrast medium (repetition time/echo time, 4/1.92 msec; flip angle, 10°; field of view, 430 mm; matrix, 192 × 138; number of sections, 37; section thickness, 5 mm; one signal acquired). For the 16 patients who underwent initial MRI examination elsewhere, the inclusion criteria were the availability of at least four sequences from a 1.5-T magnetic resonance machine, including in-phase and out-of-phase chemical shift GRE T1W images, fat-suppressed T2W images, and fat-suppressed gadolinium-enhanced T1W sequences during the arterial and delayed phases.

Image Evaluation.

All MRI data were reviewed by two abdominal radiologists (H. L., H. T.) blind to the pathological results and classification, and a consensus was obtained. In cases of multiple lesions, the largest resected HCA was analyzed. The following criteria were used for image analysis: number of lesions; location according to the Couinaud numbering system; maximum diameter; limits (clearly or poorly delimited); contour (regular or lobulated); presence of calcifications; presence of central scar; homogeneous or heterogeneous appearance of the lesion on each sequence; presence of fat deposition within the lesion (absence, focal or diffuse distribution in the lesion); presence of hemorrhagic and/or necrotic components; and presence of a peripheral enhanced rim in the delayed phase. Steatosis in non tumoral liver was evaluated on chemical shift images. Lesion intensity on each sequence was compared with the intensity of the surrounding liver parenchyma, using a four-point gray scale (hypointense, isointense, slightly hyperintense, and markedly hyperintense). On T2W images, the signal was considered markedly hyperintense if it was at least as intense as the spleen signal. If a hemorrhagic or necrotic component was found in the lesion, only the nonhemorrhagic and nonnecrotic tissue of the lesion was taken into account for signal intensity evaluation.

Statistical Analysis.

All P values < 0.05 were considered statistically significant. The Student t test was used to determine whether there was a statistically significant difference in the mean size of HCAs between the different subtypes.

We investigated the possible correlation between HCAs presenting diffuse signal dropout on chemical shift imaging (T1W opposed phase imaging) and adenoma subtype (L-FABP present or absent) using the Pearson χ2 test. This test was also used to assess the possible correlation between marked hyperintensity on T2W images and hyperintensity in the gadolinium-enhanced delayed phase on the one hand and HCA subtype (inflammatory or noninflammatory) on the other.

The sensitivity, specificity, and positive and negative predictive values of this criterion (markedly hyperintense signal on T2W images associated with hyperintense signal on delayed-phase T1W images) were determined.

Results

General Imaging Findings.

All but two of the 50 patients included were female. The mean age (± standard deviation) was 38.5 ± 7.7; 20 patients had one nodule, 22 patients had between two and 10 nodules, and eight patients had more than 10 nodules (Table 1). The mean size of the largest nodule was 64.7 ± 27 mm (Table 1). No significant difference in the size of the largest nodule was found between the HNF-1α–inactivated (67 ± 30 mm) and inflammatory (57 ± 21 mm) subtypes (P = 0.23) or between HCAs without markers (63 mm ± 32) and HNF-1α–inactivated or inflammatory HCAs (P = 0.63 and 0.73, respectively).

Hemorrhagic and/or necrotic components were identified in 10 HCAs (four inflammatory HCAs, two HNF-1α–inactivated HCAs, four HCAs without markers). The mean size of HCAs with necrosis or subacute hemorrhage was 75 mm ± 36 mm. No significant difference in mean size was found between necrotic/hemorrhagic HCAs and nonnecrotic/hemorrhagic HCAs (P = 0.15). Steatosis in the healthy parts of the liver was found in 10 patients (eight patients with inflammatory HCAs; two patients with HCAs without markers). Steatosis was never observed in liver parenchyma adjacent to HNF-1α–inactivated HCAs, whereas it was observed in eight of the 26 patients with inflammatory HCAs (P = 0.014).

In all cases of multiple nodules classified in the HNF-1α–mutated and inflammatory groups, all the nodules from the same patient tested via immunohistochemistry belonged to the same subtype. However, the intensity of SAA and CRP overproduction varied between nodules from individual patients with several inflammatory HCA lesions. In one patient with several inflammatory HCA, some of the nodules also tested positive for β-catenin (by immunohistochemistry and molecular biology), whereas others did not. We evaluated the images for two resected HCA from the same patient (patient 26) because these two tumors had similar immunohistochemical profiles but different MRI features. In all the other cases, one HCA per patient was analyzed.

In 10 cases, HCA nodules were associated with another liver tumor: typical FNH (in 4 cases), hemangioma (in 5 cases), and hepatobiliary cystadenoma (1 case). In cases of such association, we also performed immunohistochemical staining of the FNH. No SAA or CRP overproduction was found in any of the FNHs, all of which had normal L-FABP immunostaining, similar to those in the surrounding nontumoral liver.

The β-catenin–activated HCA group and HCA without markers group were very small. Therefore, the statistical analysis focused on L-FABP and inflammatory HCA. We then searched for an association between MRI features and HCA subtype.

MRI Features Associated with HNF-1α–Inactivated HCA.

Two major features were found to be significantly associated with the HNF-1α–inactivated subgroup of HCAs (Fig. 1).

Figure 1.

HNF-1α–mutated HCA. Segment IV lesion with (a) slight hypersignal on T1W images and (b) critical signal drop-out on phased opposed T1W image due to massive fat component. (c) The lesion appears slightly hyperintense on T2W with only slight enhancement after gadolinium administration in the (d) arterial and (e) portal venous phases. (f) The tumor appears yellowish on the resected specimen due to (g) diffuse steatosis. (h) Following L-FABP immunostaining, the lack of LABP in the adenoma contrasts with levels of this protein in the surrounding nontumoral liver (NTL), in which hepatocytes contain normal amounts of this protein.

First, signal dropout on chemical shift sequences, assessing the presence of fatty hepatocytes, was found in 14 of the 15 HNF-1α–inactivated HCA, but in only three inflammatory HCAs and in no β-catenin–activated HCAs or unclassified HCAs (P < 0.001). This signal dropout was diffuse and homogeneous in almost all positive HNF-1α–inactivated HCAs (13 of 14 cases), but was heterogeneous and focal in the three positive inflammatory HCAs (P < 0.0001). No cases with homogeneous signal dropout were found in β-catenin–activated and unclassified HCAs. The positive predictive value of this MRI feature (homogeneous signal dropout of the tumor on chemical shift images) for the diagnosis of HNF-1α–inactivated HCA in this population of 51 HCAs was 100%, the negative predictive value was 94.7%, the sensitivity was 86.7%, and the specificity was 100%.

Second, unlike inflammatory HCAs (see below), most HNF-1α–inactivated HCAs (13 of 15 lesions) showed only moderate arterial enhancement, and none of the lesions displayed persistent enhancement during the delayed phase.

Other MRI features (Table 2) were not found to be significantly associated with HNF-1α–inactivated HCA.

Table 2. MRI Features Considering HCA Subtype
 Group 1: L-FABP (n = 15)Group 2: SAA+ (n = 27)Group 3: β-Catenin+ (n = 2)Group 4: Unclassified (n = 7)
  1. Diameter is expressed as the mean ± standard deviation. All other data are expressed as a percentage (number of cases appears in parentheses).

  2. Abbreviation: OP, out of phase.

Diameter (mm)57 ± 2167 ± 30101 ± 1163 ± 2
Liver steatosis031 (8)029 (2)
Necrotic/hemorragic component12 (2)15 (4)071 (5)
T1W signal    
 Hypo/Iso87 (13)74 (20)100 (2)43 (3)
 Hyper13 (2)26 (7)057 (4)
OP T1W    
 Signal dropout93 (14)11 (3)014 (1)
 Homogeneous repartition of signal dropout87 (13)000
T2W signal    
 Homogenous87 (13)44 (12)00
 Hypo/Iso47 (7)0100 (2)57 (4)
 Hyper53 (8)15 (4)043 (3)
 Markedly hyperintense085 (23)Focal: 100 (2)0
 High signal peripheral rim081 (22)00
Enhancement    
 Strong arterial enhancement13 (2)93 (25)100 (2)71 (5)
 Delayed wash-out00100 (2)0
 Persistent enhancement089 (24)Focal: 100 (2)Focal: 29 (2)

MRI Features Associated with Inflammatory HCA.

Inflammatory HCA was significantly associated with the following MRI features (Figs. 2 and 3).

Figure 2.

Inflammatory HCA. (a) The 60-mm lesion located in segment VII is isointense with the surrounding parenchyma on T1W images with (b) no signal drop-out on chemical shift sequence. The lesion presents a high signal intensity on both (c) T2W and (d) T2W fat-suppressed images. (e) Strong arterial enhancement after gadolinium administration, and (f) persistent enhancement in the portal venous (not shown) and delayed phases 5 minutes after gadolinium-based contrast injection. (g) The hepatocellular tumor is clearly telangiectatic, with (h) inflammatory infiltrates (arrow) often surrounding thick arteries (asterisk). (i) Adenomatous hepatocytes overexpressed SAA, with clear delimitation between adenoma and nontumoral liver (NTL).

Figure 3.

Inflammatory HCA. (a) The 10-cm lesion located in segment VII is isointense with the surrounding parenchyma on T1W images with no signal drop-out on chemical shift sequence (not shown). (b) The periphery of the lesion presents a high signal intensity on T2W fat-suppressed images, whereas the signal is weaker in the central part, with small intratumoral hemorragic zones (arrowheads). A strong correlation was found between the zones of intense signal on T2W images and telangiectatic areas. (c) Strong arterial enhancement was observed after gadolinium administration, with persistent enhancement in the (d) portal venous and (e) delayed phase, after 5 minutes. (f) On light microscopy, the tumor clearly displayed telangiectasia at the periphery (asterisk), whereas the center of the tumor was more compact. (g) Inflammatory infiltrates (arrow) are observed around arteries with thickened walls. (h) On SAA immunostaining, adenomatous hepatocytes overexpressed SAA, whereas inflammatory cells tested negative for this protein.

A hyperintense signal on T2W images was found in all inflammatory HCAs (27 analyzed cases) but in only eight HNF-1α–inactivated HCAs (P < 0.001). The signal was markedly hyperintense in 23 of the 27 inflammatory HCAs, but in none of the HNF-1α–inactivated HCAs (P < 0.001). A target feature with a stronger signal in the outer part of the lesion was found in 22 inflammatory HCAs, but in none of the HNF-1α–inactivated HCAs (Fig. 3).

Strong arterial enhancement was found in 25 of 27 inflammatory HCAs and in only two HNF-1α–inactivated HCAs (P < 0.001).

Persistent enhancement in the delayed phase was found in 24 of the 25 positive inflammatory HCAs.

An association of markedly hyperintense signal on T2W images and persistent enhancement on delayed-phase T1W images was found in 23 of 27 inflammatory HCAs, but in only three HNF-1α–inactivated HCAs and three unclassified HCAs (P < 0.0001). The positive predictive value of these two MRI features for the diagnosis of inflammatory telangiectatic HCA in this population of 51 HCAs was 88.5%, the negative predictive value was 84%, the sensitivity was 85.2%, and the specificity was 87.5%.

Other MRI features (Table 2) were not found to be significantly associated with inflammatory HCA.

MRI Features with β-Catenin Activation.

Five HCAs displayed β-catenin activation. Three of these five tumors also displayed inflammation. These three nodules displayed the MRI pattern typical of inflammatory HCAs (markedly hyperintense signal on T2W images, strong arterial enhancement and persistent enhancement in the delayed phase). The first of the three patients concerned had one nodule, the second had five nodules, and the third had 10 nodules.

The two noninflammatory β-catenin–activated HCAs had similar MRI features (Fig. 4). Both lesions were heterogeneous on all sequences, but no necrotic or hemorrhagic component was found. There was no signal dropout on out-of-phase chemical shift sequences. Most of the tumor was isointense on T1W and T2W images, with strong arterial enhancement and delayed wash-out after gadolinium injection. The rest of the tumor resembled inflammatory HCA, with marked hyperintensity on T2W sequences and persistent delayed enhancement after gadolinium administration.

Figure 4.

Unclassified HCA. (a) The segment VII tumor displays a slightly hyperintense signal (arrowheads) on T1W images and (b) an isointense signal with respect to the liver parenchyma on T2W images. (c) Strong enhancement is observed in the arterial phase, with (d) tumor signal isointensity during the delayed phase. (e) On HES, the tumor (HCA) displays no particular features, with no steatosis, inflammation or cytological abnormality in particular.

MRI Features Associated with Unclassified HCA.

In HCAs without specific markers, four lesions had the characteristics of benign hepatocellular tumors with no specific MRI features (Fig. 5). Two of these tumors contained hemorrhagic zones, and one had a focal fat component. Three lesions were heterogeneous in appearance on all sequences, with a strong central signal on T2W images, related to areas of necrosis or peliosis. The MRI characteristics of the peripheral areas of these three lesions was typical of hepatocellular tumors (isointense signal on T1W and T2W images, strong arterial enhancement with no delayed enhancement after gadolinium injection).

Figure 5.

β-Catenin–mutated HCA. (a) The segment VII lesion is slightly hypointense with respect to the liver parenchyma on T1W images. (b) On T2W images, most of the tumor appears to be isointense, with a poorly delimited high-intensity area (arrowheads). (c) After gadolinium administration, strong heterogeneous enhancement was observed in the arterial phase, with wash-out of most of the lesion in the (d) portal and (e) delayed phases. Persistent enhancement was observed during the portal and delayed phases in the poorly delimited area of signal hyperintensity on T2W images. (f) Macroscopic examination of fresh tissue showed the tumor to be isochromic. (g) No specific pathological feature (no steatosis, no inflammation, thin vessels) was identified on hematoxylin-eosin staining, but there were numerous anastomosing vessels at the periphery of the tumor, with poor delimitation between the adenoma and nontumoral liver (NTL). (h) In the area with numerous vessels, sinusoids were capillarized as shown on CD34 staining.

MRI Features Associated with Multiple Nodules.

Multiple nodules were found in 17 patients with inflammatory HCAs, 11 patients with HNF-1α–inactivated HCAs, no patient with noninflammatory β-catenin–activated HCAs, and three patients with unclassified HCAs. In all patients other than one patient with inflammatory HCAs, the nodules from an individual patient shared the same MRI pattern. Three patients with inflammatory HCAs and six patients with HNF-1α–inactivated HCAs fulfilled the criteria for liver adenomatosis, with more than 10 nodules detected on MRI scans. In all cases, liver adenomatosis involved multiple hepatic segments with no right or left lobe predominance.

Discussion

In this study, we identified specific MRI features closely associated with HNF-1α–inactivated or inflammatory HCA—the two most frequently identified subtypes of HCA.18

Because MRI features were similar in different nodules from the same patient with multiple HCAs—apart from occasional areas of hemorrhagic necrosis, differing from one nodule to another—we analyzed the excised nodule if only one nodule was removed or the largest excised nodule if several were removed.

This study shows that HNF-1α–inactivated HCA can be recognized with confidence on MRI if a homogeneous fat distribution is observed, particularly on chemical shift sequences. Pathological reviews of large series of HCA have shown marked steatosis to be a feature common to almost all cases of HNF-1α–inactivated HCA.3, 4 This steatosis results from the activation of lipogenesis.5 Steatosis is also observed in other subtypes of HCA, but less frequently and to a lesser degree. Thus, no lesion without HNF-1α inactivation shows such a diffuse distribution of fat, although foci of steatosis may be observed in inflammatory HCA. Furthermore, enhancement after gadolinium-based contrast injection displayed a particular pattern, with only two tumors showing strong arterial enhancement and none showing persistent enhancement in the delayed phase.

Inflammatory HCA cases (formerly known as telangiectatic FNH19, 20) also displayed rather specific MRI features. The characteristic MRI pattern of this subgroup included (1) a markedly hyperintense signal on T2W images, (2) strong arterial enhancement, and (3) persistent enhancement in the delayed phase. The combination of a markedly hyperintense signal on T2W images and persistent enhancement in the delayed phase were found to be sensitive (85.2%) and specific (87.5%) for the radiological diagnosis of inflammatory HCA in this population of 50 HCAs. A target-like feature was observed for a large proportion of these lesions on T2W images, with a very high-intensity signal forming a rim in the outer part of the lesion. A good correlation was found between these rims and the telangiectatic area (Fig. 3). We hypothesized that the above MRI features of inflammatory HCA are related to the pathologic characteristics of these lesions; indeed, the presence of numerous arteries, of major sinusoidal dilatation usually at their periphery and of more or less numerous and large cavities, indicate major hemodynamic disturbances. More specifically, it is possible that the sinusoidal dilatation at the periphery of inflammatory HCA could induce an hyperintense signal on T2W images by slowing blood flow and an hyperintense signal in the equilibrium late phase by accumulation of contrast medium. Signal intensity on T1W images does not seem to be specific enough for the identification of inflammatory HCsA, because a large proportion of both inflammatory and HNF-1α–inactivated HCAs showed isosignal or hyposignal intensity. No lesion showed diffuse signal dropout on the chemical shift sequence, but focal signal dropout was found in 11% (n = 3) of the tumors in this group related to patchy fat deposition. These three tumors had the other specific MRI features of inflammatory HCA, including a markedly hyperintense signal on T2W images and persistent enhancement in the delayed phase.

Three HCAs were both inflammatory and displayed β-catenin activation and could not be distinguished using pathological methods or MRI from other cases of inflammatory HCA. Further study of inflammatory HCA with β-catenin activation is needed to confirm this observation.

The two cases with β-catenin activation but with no inflammation or features consistent with their classification as HCC or borderline lesions were identical. They contained areas with numerous veins, correlated with a hypersignal on T2W images and a persistent enhancement area on MRI, with most of the lesion isointense on T2W and displaying strong arterial enhancement with marked wash-out at the portal venous phase and during the delayed phase (Fig. 5). This feature (hypointensity of the signal in the delayed phase) has already been reported in HCA10 but is best known as a characteristic MRI feature of malignant lesions such as HCC. Furthermore, no other lesion in this study showed delayed wash-out. It is clear that additional studies are needed to confirm these findings; it is still too early to propose MRI as a method of identification for this HCA group.

Finally, no specific pathological or immunohistochemical criteria were identified for the seven unclassified cases. We identified no useful MRI criteria for the diagnosis of this group. Some of these tumors were heterogeneous in appearance due to areas of necrosis or bleeding (n = 5), the remaining tissue of the lesion showing the classical MRI characteristics of benign hepatocellular lesions. Areas of peliosis were found in two lesions and correlated with hypersignal on T2W images and persistent enhancement on MRI. These findings did not overlap with the characteristic MRI pattern of inflammatory HCA, because only a small central part of the tumor was concerned.

Beaujon's group in Clichy21 has shown that MRI can be used to identify different groups of HCA: steatotic, peliotic, and mixed. It seems reasonable to assume that the steatotic group in their small series corresponds to our HNF-1α–inactivated group. However, because the correlation between peliosis (sinusoidal dilation) and inflammation is not strict, it is unclear whether Beaujon's peliotic group is equivalent to our cases of inflammatory HCA. Unlike Beaujon's group, we did not identify a mixed steatotic and peliotic form consisting of large lesions with steatotic areas, hypersignal on T2W images, and enhancement in the arterial, portal, and delayed phases. As indicated above, steatosis may be observed in inflammatory HCA. This focally and irregularly distributed steatosis clearly differs from the diffuse distribution of fat in HNF-1α–inactivated HCA. We have also not observed cases combining a lack of L-FABP with the presence of SAA on immunohistochemical staining. We therefore believe that the presence of a focal fat component should not call into doubt the MRI-based diagnosis of inflammatory HCA when other MRI criteria are present. Lesions of this type showing both focal areas of steatosis and MRI features characteristic of inflammatory lesions probably correspond largely to Beaujon's mixed form lesions.

This study demonstrates a good relationship between MRI data and genotype/phenotype classification for the two major subtypes. We have already shown that clinical and biochemical data may differ between HNF-1α–mutated HCA and inflammatory HCA. We hope that by combining clinical, biological, and MRI data it will be possible to improve the classification of these two types of HCA (which account for 80% of all cases) even further. It would be useful to determine whether these two major subtypes behave differently in terms of their natural history (growth, involution following the cessation of oral contraceptive use or at menopause) and their complications, such as the risk of hemorrhage or malignancy, regardless of size.

One major limitation of this study is that MRI cannot yet differentiate between the SAA group and the SAA+ β-catenin–mutated group. Liver biopsy remains the only available tool for evaluation of these two groups. However, liver biopsy is probably no longer useful for confirming diagnosis of HNF-1α–inactivated HCA, a subtype known to be associated with a lower risk of malignant transformation.

The next step will be the validation of these data for an independent group of unselected patients with liver nodules, determining whether the criteria defined above can differentiate not only between the two major subgroups of HCA, but also between HCA and other benign or malignant primary nodules. Multicenter studies will be required to address this issue, given the rarity of HCA. We hope that, by combining additional genotypic/phenotypic markers, MRI, and contrast-enhanced ultrasound scans, it may be possible to improve the identification of HCA subtypes in a noninvasive manner.

In conclusion, MRI is a useful tool for identifying the two major subtypes of HCA (HNF-1α and β-catenin mutations), which have distinctly different patterns on MRI.

Acknowledgements

We thank Professor Jean Saric and Professor Antonio Sa Cunha (surgeons) and Professor Brigitte Le Bail and Dr. Anne Rullier (liver pathologists) for their clinical contributions to this study.

Ancillary