Gadoxetic acid (gadolinium ethoxybenzyl diethylenetriaminepentaacetic acid; Gd-EOB-DTPA; Primovist ® (sold as Eovist ® in the USA), Bayer Schering Pharma AG, Berlin, Germany) is a contrast agent for magnetic resonance (MR) imaging with hepatocyte-specific property [1, 2]. Gadoxetic acid-enhanced MRI (EOB-MRI) has been becoming more important in the detection and characterization of hepatic mass lesions including hepatocellular carcinoma (HCC) and other hepatic nodules [3,4,5,6,7,8].

Recent studies have revealed the transporter mechanism of gadoxetic acid in HCCs. Sinusoidal membranous uptake transporter organic anion-transporting polypeptide (OATP)1B3 (synonymous with OATP8) expression in HCC cells determines the signal intensity of HCC on the hepatobiliary (HB) phase of EOB-MRI [9, 10]. In addition, benign hepatocellular nodules including hepatocellular adenoma (HCA), focal nodular hyperplasia, and dysplastic nodule also show a positive correlation between signal intensity on HB phase of EOB-MRI and OATP1B3 expression [11,12,13,14,15,16]. Therefore, OATP1B3 is thought to be the main uptake transporter of gadoxetic acid determining the signal intensity on HB phase in hepatocellular nodules including hyperplastic lesions.

In addition, the molecular background of OATP1B3 expression has been becoming clearer. In normal liver parenchyma, OATP1B3 and glutamine synthetase (GS), the downstream target of the beta-catenin signaling, are considered to be expressed by the interaction of Wnt/β-catenin signaling and hepatocyte nuclear factor (HNF) 4α in peri-central (zone 3) hepatocytes [17,18,19,20]. Therefore, co-expression of GS and OATP1B3 is seen in peri-central hepatocytes compared with hepatocytes surrounding the portal venule [17,18,19,20]. In HCC, an intense correlation between beta-catenin mutation with GS expression and OATP1B3 expression was also reported [21, 22], and HCC showing hyperintensity relative to the surrounding liver (OATP1B3 over-expressed HCC) on HB phase was confirmed to be associated with activation of β-catenin signaling, GS, and HNF4α [23, 24]. Moreover, β-catenin-activated HCA shows over-expression of GS and OATP1B3 resulting in iso/hyperintensity on HB phase of EOB-MRI [12, 13]. Therefore, an intense correlation between OATP1B3, which is the most important determinant of Gd-EOB-DTPA uptake into hepatocytes and/or HCC cells, and GS expression has been strongly suggested.

In our experience, peri-tumoral hyperintensity (P-hyperintensity) surrounding a HCC nodule is occasionally seen on HB phase of EOB-MRI. However, the mechanism of P-hyperintensity is still unknown. On the other hand, a recent study showed peri-tumoral hyperplasia (P-hyperplasia) of hepatocytes associated with over-expression of GS surrounding HCC or metastatic carcinoma [25]. It was defined as a peri-tumoral rim of hyperplastic hepatocytes with altered plate architecture (including hepatocyte rosettes or thickened plates) delimited by a region of parenchymal atrophy and with diffuse hepatocellular expression of GS and variable sinusoidal CD34 expression, which was surmised to be due to peri-tumoral portal blood flow obstruction.

Based on these facts and previous reports, we speculated that the P-hyperintensity on HB phase of EOB-MRI might be due to P-hyperplasia with higher expression of GS and OATP1B3 and might have some relation with the biological nature of HCC.

The aim of this study was to analyze the correlation between P-hyperintensity on HB phase of EOB-MRI and immunohistochemical expression of GS indicating P-hyperplasia in the liver parenchyma surrounding HCC, and its relation with pathological features of HCCs.

Materials and methods

Patients

This retrospective study received the approval of our institutional ethics committee. Firstly, we included 86 consecutive surgically resected patients confirmed as HCC pathologically in our hospital during the period from 2008 to 2010. And we excluded 16 patients because EOB-MRI was not performed. In addition, two patients were excluded because of poor quality of HB phase image. We finally included 77 HCCs from 68 patients in our hospital during the period from 2008 to 2010. Mean patient age was 65.2 years ± 9.73 (standard deviation) (range 38–81 years) for the entire study group. The male/female ratio was 53:24, and 68 of 77 HCCs had a background of chronic hepatitis or liver cirrhosis. The level of serum AFP was the following: normal level of serum AFP (< 10 ng/mL) (n = 23), 10–100 ng/mL (n = 29), more than 100 ng/mL (n = 25). And Child–pugh score was the following: Child–pugh score A (n = 71), B (n = 6) (Table 1).

Table 1 Summary of the patients
Full size table

Imaging details

MR images were obtained with a 1.5/3.0 Tesla MR system (Signa HDx; GE Medical Systems, Milwaukee, WI). The protocols for T1-weighted, T2-weighted, diffusion-weighted imaging and contrast material-enhanced T1-weighted imaging are shown in Table 2. T1-weighted imaging was performed during precontrast and hepatobiliary phase. T2-weighted imaging was obtained after the dynamic study. A dose of 0.1 mL/kg of gadoxetic acid was intravenously injected at a flow rate of 1.0 mL/s, followed by a 30-mL saline flush. We employed the test injection method (1.5 mL of gadoxetic acid plus an 8-mL saline flush) to determine the optimal arterial dominant phase, which was determined as the peak time of enhancement in the abdominal aorta plus an additional 10 s and half of imaging time in the arterial phase. Portal phase and equilibrium phase images were obtained at 60–90 s and 120–180 s after injection, respectively. The HB phase was obtained 20 min after the injection in all cases. The mean interval between surgery and EOB-MRI was 50.3 days.

Table 2 Protocol for MR imaging
Full size table

Contrast-enhanced dynamic CT images were obtained with CT systems (VCT64; GE Medical Systems, Milwaukee, Wis, or Somatom Definition Flash; Siemens, Erlangen, Germany) with the following parameters: voltage, 120 kV; amperage, 300–350 mA; rotation period, 0.5 s; detector collimation, 0.6–0.625 mm; pitch, 0.516–1.0; and section thickness, 2.5–3.0 mm. 100–150 mL of contrast material with 600 mg of iodine per kilogram (Iopamiron; Bayer Schering Pharma, Berlin; Iomeron; Eisai, Tokyo, Japan; Omnipaque; Daiichi-Sankyo, Tokyo, Japan; or Optiray; Mallinckrodt Japan, Tokyo, Japan) was injected for 30 s after precontrast scanning. The equilibrium phase was obtained after a scanning delay of 150 s. The mean interval between surgery and dynamic CT was 65.7 days.

Imaging analysis

We just defined P-hyperintensity as hyperintensity rim surrounding the tumor. On HB phase of EOB-MRI, the fibrous capsule is not evaluated clearly. If there is a hypointense rim surrounding the tumor suggesting a fibrous capsule, we evaluated the outside of the capsule like rim when we evaluated P-hyperintensity.

HCCs were grouped according to the degree of the peri-tumoral hyperintense signal at the level of the largest diameter of the tumor on the HB phase of EOB-MRI: grade 0 (no P-hyperintensity), grade 1 (P-hyperintensity was seen on less than 50% of the tumor border), grade 2 (50%–80%), grade 3 (80%–100%) (Fig. 1). We also evaluated the presence of peri-tumoral hypointensity surrounding the tumor on HB phase of EOB-MRI as reported previously [26].

Fig. 1
figure 1

P-hyperintensity grading on HB phase of EOB-MRI. A Patients were grouped according to the degree of the peri-tumoral hyperintense signal on the HB phase of EOB-MRI: grade 0 (no P-hyperintensity), grade 1 (P-hyperintensity was seen on less than 50% of the tumor border), grade 2 (50%–80%), grade 3 (80%–100%). B. Grade 0 (no P-hyperintensity), there is no P-hyperintensity around a HCC on HB phase. C. Grade 1, P-hyperintensity was seen on less than 50% of the tumor border. D. Grade 2, P-hyperintensity was observed on less than 50% of the tumor border. E. Grade 3, P-hyperintensity was observed on more than 80% of the tumor border

Full size image

We evaluated HCC imaging features including enhancement pattern and signal intensity on T1-weighted, T2-weighted, and diffusion-weighted imaging. As an enhancement pattern, we defined a pattern that shows arterial enhancement and washout at equilibrium phase as a typical enhancement pattern for HCC. We evaluated the arterial enhancement of the nodule on arterial dominant phase of EOB-MRI. In this analysis, hypovascular nodule was defined as a nodule showing hypointensity compared with surrounding liver on arterial dominant phase of EOB-MRI.

In the area of peri-tumoral corona enhancement on dynamic CT due to the drainage flow from hypervascular HCC, portal flow is decreased [27]. Therefore, this may have some relation to P-hyperplasia. We evaluated corona enhancement on equilibrium phase by dynamic CT. Because the transitional phase of EOB-MRI was not suitable for evaluation of corona enhancement due to the effect of gadoxetic acid uptake in the surrounding liver parenchyma. Three cases did not undergo preoperative dynamic CT imaging. Therefore, a total of 74 lesions were subjected to the analysis of corona enhancement. Corona enhancement was divided into the following three degrees: clear corona enhancement at the thickest part 3 mm or more, thin corona enhancement, and no corona enhancement.

Image analysis was performed independently by two abdominal radiologists (N.Y. with 15 years of experience and T.K. with 6 years of experience) who were blinded to the histological existence of P-hyperplasia in the nodule. When the decisions of the two radiologists differed, the final evaluation was made by consensus.

Histologic diagnosis

HCCs were diagnosed by a liver pathologist (H.I. with 16 years of experience) according to the classification proposed by the international Working Party [28], WHO classification [29], and general rules proposed by the liver cancer study group of Japan [30]. We analyzed histologic features such as the presence or absence of fibrous capsule, differentiation grade (well, moderately, and poorly differentiated), macroscopic growth patterns (indistinct margin, simple nodular, nodular type with extranodular growth, confluent multinodular infiltrative patterns), and vascular invasion including microscopic portal vein invasion and hepatic venous invasion. And liver fibrosis and inflammation of background liver was graded according to the Inuyama criteria [31]: no fibrosis (F0), portal fibrosis widening (F1), portal fibrosis widening with bridging fibrosis (F2), bridging fibrosis and lobular distortion (F3), cirrhosis (F4), none (A0), mild (A1), moderate (A2), and severe (A3) inflammation.

Immunohistochemistry

The specimens were fixed in 10% formalin after surgical removal. Immunohistochemical stainings were carried out in all nodules using the primary antibodies against human GS (mouse monoclonal, GS-6; Millipore, Billerica, Mass) and human OATP1B3 (mouse monoclonal, NB100-74482; Novus Biologicals, Littleton, Colo). One representative specimen per nodule containing the boundary between tumor and background liver was subjected to the analysis. We could not evaluate the entire tumor circumference in this study.

Definition of P-hyperplasia

Previous study revealed that P-hyperplasia lesion has diffuse immunopositivity for GS [24]. GS is present normally in hepatocytes adjacent to the hepatic venule. In this study, P-hyperplasia was defined as a rim of hyperplastic hepatocytes at least 0.5 mm thickness surrounding the tumor with higher GS expression as compared with those in non-peri-tumoral liver (Fig. 2). The membranous expression of OATP1B3 in the hepatocytes in P-hyperplasia was also evaluated as compared with that in non-peri-tumoral background liver. Immunohistochemical analysis was performed by two abdominal radiologists (N.Y. with 15 years and A.K. with 16 years of experience, both of whom are graduates of the Department of Pathology, Kanazawa University, in addition to being board certificated radiologists). When the decisions of the two observers differed, the final decision was made by consensus.

Fig. 2
figure 2

GS expression of P-hyperplasia. P-hyperplasia was defined as hyperplastic peri-tumoral hepatocytes with higher GS expression as compared with those in non-peri-tumoral liver in this study. A P-hyperplasia negative, P-hyperplasia with higher GS expression is not observed in the peri-tumoral area. (Original magnification, ×20). B P-hyperplasia positive, P-hyperplasia with higher GS expression is seen in the peri-tumoral area. (Original magnification, ×40)

Full size image

Statistical analysis

Statistical significance was evaluated with software (Prism5; GraphPad Software, San Diego, Ca). The χ 2 test, Fisher’s exact test, and Mann–Whitney U test were used for the analysis of the relationship between P-hyperplasia and P-hyperintensity grade, peri-tumoral hypointensity, and their imaging and pathological analysis. P < 0.05 was considered to indicate a statistically significant difference. Inter-reader agreement was assessed by calculating a κ value or weighted κ value which was interpreted as follows: poor agreement, ≤ 0.20; fair agreement, 0.21–0.40; moderate agreement, 0.41–0.60; good agreement, 0.61–0.80; and excellent agreement, 0.81–1.00.

Results

Imaging diagnosis of HCC

In this study, 72 HCCs from 77 HCCs demonstrated typical enhancement pattern (arterial enhancement- washout) as HCC and other 5 HCCs had atypical enhancement pattern (hypovascular). 72 HCCs showing typical enhancement pattern could be diagnosed as HCC by imaging. Other imaging features of HCCs including signal intensity on T2-weighted, T1-weighted and diffusion-weighted imaging are shown in Table 3. Of 72 typical enhancement HCCs, 2 HCCs showed partial hyperintensity and 5 HCCs showed hyperintensity on HB phase of EOB-MRI. On the other hand, all atypical enhancement HCCs showed hypointensity on HB phase. Therefore, we diagnosed those atypical enhancement groups as hypovascular HCC by imaging preoperatively.

Table 3 Imaging findings of HCC
Full size table

Frequency of the P-hyperintensity

In the image evaluation of the HB phase of EOB-MRI, 34 nodules (44.2%) were classified as grade 0, 29 nodules (37.7%) as grade 1,10 nodules (13.0%) as grade 2, and 4 nodules (5.2%) as grade 3. And the peri-tumoral hypointensity was observed in 27/77 (35.1%) HCCs. 18 HCCs had both P-hyperintensity and peri-tumoral hypointensity on HB phase of EOB-MRI. Neither P-hyperintensity nor peri-tumoral hypointensity was seen in 25 HCCs (32.5%). The evaluation of the P-hyperintensity grade differed between the two readers for 7 HCCs (9.1%). With visual evaluation, the weighted kappa value of the P-hyperintensity grade became 0.825. Of seven evaluation differences, 5 occurred between grades 0 and 1 and 2 occurred between grades 2 and 3 among two readers. Regarding the 5 differences of between grade 0–1, it was caused by the difference whether to take subtle findings of P-hyperintensity or not. And two differences between grades 2 and 3 occurred due to the difference in interpretation of proportion of P-hyperintensity.

Histological assessment of P-hyperplasia and correlation between P-hyperintensity and P-hyperplasia

Histologically total 32 lesions (41.6%) had P-hyperplasia showing high GS expression. The thickness of P-hyperplasia ranged from a maximum of 2.2 mm to a minimum of 0.5 mm. The relationship between P-hyperplasia and P-hyperintensity grade is shown in Fig. 3. P-hyperplasia was observed in 3/34 P-hyperintensity grade 0 nodules (8.8%), 16/29 grade 1 nodules (55.2%), 9/10 grade 2 nodules (90%), and 4/4 grade 3 nodules (100%) (Figs. 3, 4). P-hyperplasia was more frequently observed in higher P-hyperintensity grade nodules on the HB phase of EOB-MRI. There was a significant difference in the incidence of P-hyperplasia between P-hyperintensity grade 0 group and grade 1–3 group (p < 0.0001). The negative predictive value of this sign predicting P-hyperplasia was 91.2 (31/34) %. Hepatocytes with high GS expression indicating P-hyperplasia demonstrated definite membranous expression of OATP1B3 (uptake transporter of EOB) in all P-hyperplasia sites. In the normal liver, OATP1B3 expression shows zonation and is expressed predominantly in the centrilobular hepatocytes (zone 3) [32]. In P-hyperplasia sites, diffuse OATP1B3 expression was observed and the zonation of OATP1B3 expression was lost, differing from the pattern seen in non-peri-tumoral liver (Fig. 5).

Fig. 3
figure 3

Frequency of P-hyperplasia according to P-hyperintensity grade on EOB-MRI. P-hyperplasia was more frequently observed in higher P-hyperintensity grade nodules on HB phase of EOB-MRI. There was a highly significant correlation in the incidence of P-hyperplasia between P-hyperintensity grade 0 group and grade 1–3 group (p < 0.0001)

Full size image
Fig. 4
figure 4

Representative lesions of P-hyperintensity grade 0 without P-hyperplasia and P-hyperintensity grade 2 with P-hyperplasia. Representative lesions of P-hyperintensity grade 0 without P-hyperplasia (A, B) and P-hyperintensity grade 2 with P-hyperplasia (C, D). A This lesion has no P-hyperintensity on HB phase of EOB-MRI. B GS staining shows no hepatocytes with high GS expression indicating P-hyperplasia. (Original magnification ×20). C This lesion was evaluated as P-hyperintensity grade 2. D GS staining shows definite peri-tumoral hepatocytes with higher GS expression (P-hyperplasia). (Original magnification ×20)

Full size image
Fig. 5
figure 5

P-hyperintensity grade 3 shows P-hyperplasia with over-expression of OATP1B3. A HB phase shows peri-tumoral hyperintensity around almost the whole boundary. This lesion was evaluated as P-hyperintensity grade 3. B GS staining shows definite P-hyperplasia. (Original magnification ×100). C In HE staining, P-hyperplasia can be distinguished from the background liver (Original magnification ×100). D High magnification of HE staining shows hyperplastic hepatocytes with no atypia at P-hyperplasia site. (Original magnification ×200) E Hepatocytes with GS expression indicating P-hyperplasia demonstrated definite membranous expression of OATP1B3 (uptake transporter of EOB) in all P-hyperplasia areas. Zonation of OATP1B3 expression was lost at P-hyperplasia sites. (Original magnification ×400) F In non-peri-tumoral liver parenchyma, OATP1B3 are expressed predominantly in the centrilobular hepatocytes (zone 3). (Original magnification ×400)

Full size image

Correlation between the grade of P-hyperintensity and imaging and pathological features

The correlation between the presence of P-hyperintensity (grade 0 or grades 1–3) and imaging and pathological features was also analyzed (Table 4). There were no significant differences in the following factors including sex, age, size, hypovascular nodule on arterial phase, degree of corona enhancement, presence or absence of fibrous capsule, histology, macroscopic classification, and background liver fibrosis and inflammation. With regard to microvascular invasion in P-hyperintensity-positive HCCs (grades 1–3), the incidence of microscopic portal vein invasion, microscopic hepatic venous invasion, and microscopic vascular invasion (portal or hepatic venous) was seen in 20/43 (46.5%), 11/43 (25.6%), and 23/43 (53.3%), respectively. There was no significant difference in microscopic portal vein invasion (p = 0.4662), and microscopic vascular invasion (portal or hepatic venous) (p = 0.1828) among P-hyperintensity-positive (grade 1–3) and P-hyperintensity-negative (grade 0) HCCs. On the other hand, microscopic hepatic venous invasion was significantly more often seen in P-hyperintensity-positive group (grades 1–3) (p = 0.0017).

Table 4 Imaging and pathological analysis among P-hyperintensity-negative (grade 0) and P-hyperintensity-positive (grade 1–3) group
Full size table

In addition, in this study peri-tumoral hypointensity was observed in 27 HCCs. In HCCs with peri-tumoral hypointensity, microscopic portal vein invasion, microscopic hepatic venous invasion and microscopic vascular invasion (portal or hepatic venous) were seen in 18/27 HCCs (66.7%), 8/27 (29.6%), and 20/27 HCCs (74.1%), respectively. And there was a significant difference in microscopic portal vein invasion (p = 0.0034), microscopic hepatic venous invasion (p = 0.0130) and microscopic vascular invasion (portal or hepatic venous) (p = 0.0007) between peri-tumoral hypointensity-positive HCCs (n = 27) and peri-tumoral hypointensity-negative HCCs (n = 50).

Between the two readers, the evaluation of the degree of corona enhancement differed for 5 HCCs (6.5%) and of hypovascularity in the arterial phase for 1 HCC (1.3%). Regarding inter-reader reproducibility, the weighted kappa value of the degree of corona enhancement and kappa value of hypovascularity were 0.909 and 0.902, respectively.

Discussion

In this study, there was a significant positive correlation between the presence of P-hyperplasia with over-expression of GS and the grade of P-hyperintensity. In addition, P-hyperplasia sites showed definite OATP1B3 (uptake transporter of Gd-EOB-DTPA) expression. Therefore, P-hyperintensity on the HB phase of EOB-MRI can be considered to reflect P-hyperplasia in the majority of lesions. The sensitivity and specificity of the P-hyperintensity predicting the presence of P-hyperplasia was 90.6% and 67.4%. We think that the reason for the relatively low specificity of this sign was due to the ambiguity of the finding of the grade 1.

As mentioned in the introduction, OATP1B3 and GS gene in peri-central hepatocytes are speculated to be regulated by Wnt/beta-catenin signaling [17,18,19,20]. In addition, in HCC and HCA a close relationship between OATP1B3 expression and Wnt/beta-catenin signaling including GS has been reported [12, 13, 21,22,23,24]. Therefore, GS might be closely related to OATP1B3 expression. In fact, OATP1B3 was well expressed at all P-hyperplasia sites with over-expression of GS in this study. The present result showing that p-hyperplasia with marked expression of GS has definite OATP1B3 expression can be explained by these molecular mechanisms. This might contribute to the hyperintensity seen on the HB phase of EOB-MRI.

As mentioned above, P-hyperplasia was diagnosed if there was a rim of hyperplastic hepatocytes of at least 0.5 mm thickness with plate architecture (including hepatocyte rosettes or thickened plates), delimited by a region of parenchymal atrophy with GS expression. In this study, P-hyperplasia was defined only by the presence of peri-tumoral hepatocytes rim with high GS expression. This is one of the important limitations of this study, and further pathological definition of this phenomenon is needed.

In this study, P-hyperintensity grade 1 had a lower incidence (55.2%) of histological P-hyperplasia compared with P-hyperintensity grade 2 nodules (90%) and 3 nodules (100%). This might be attributable to a misreading of artifacts or adjacent fine bile ducts as P-hyperintensity. Moreover, we did not evaluate the entire circumference histologically in this study, with this possibly being another cause of the lower incidence of histological P-hyperplasia noted in P-hyperintensity grade 1. In addition, fibrous capsule is thought to be the other possible cause of P-hyperintensity due to the retention of contrast media; however, in this study, the presence of fibrous capsule did not correlate with P-hyperintensity. Therefore, fibrous capsule may not be the main cause of P-hyperintensity. And also, the background fibrosis and inflammation also did not correlate with P-hyperintensity.

The pathogenesis of P-hyperplasia remains unclear. The following three possible causes have been considered: (1) perfusion abnormalities (due to vascular invasion), (2) regenerative changes in the liver cells induced by tumor compression, and (3) association with drainage blood flow. As for (1), Arnason et al. [25] suggested that P-hyperplasia is a hyperplastic response to increased arterial blood flow in the peri-tumoral parenchyma due to portal vein invasion. However, in this study, the frequency of portal vein invasion was not significantly increased in P-hyperintensity-positive (grade 1–3) group compared with negative of P-hyperintensity (grade 0). On the other hand, microscopic hepatic venous invasion was significantly more often seen in P-hyperintensity-positive (grade 1–3) group. Therefore, perfusion abnormalities due to microscopic hepatic venous invasion might be one cause of P-hyperintensity and P-hyperplasia. As for (2), GS has been also reported to be expressed in progenitor cells of the liver [33]. Therefore, regenerative changes may be one cause of P-hyperplasia. Concerning about (3), in HCC, corona enhancement due to drainage flow from the tumor into the surrounding liver is commonly observed. This drainage area does not receive portal blood flow and may be a cause of P-hyperplasia. However, in our study a close relationship was not noted between the degree of corona enhancement and P-hyperintensity grade. Therefore, drainage flow from the tumor may not directly contribute to the formation of P-hyperplasia.

In this study, although there was a statistically significant difference in microscopic hepatic vascular invasion between P-hyperintensity-positive group (grades 1–3) and P-hyperintensity-negative group (grade 0) (p = 0.0017), the incidence of microscopic hepatic venous invasion in P-hyperintensity-positive HCCs (grades 1–3) was relatively low (11/43; 25.6%). Therefore, we consider that the clinical impact of P-hyperintensity predicting microscopic hepatic venous invasion is thought to be limited.

On the other hand, previous study showed peri-tumoral hypointensity was correlated with higher incidence of microscopic vascular invasion. Kim et al. reported the incidence of microvascular invasion in HCC with peri-tumoral hypointensity was 88.5% [26]. In our study, HCC with peri-tumoral hypointensity was observed in 27/77 (35%) and the incidence of microvascular invasion (hepatic venous or portal venous) was 74.1%. Although the incidence rate of microvascular invasion in HCC with peri-tumoral hypointensity in our study was a little less frequent, a high incidence of microvascular invasion (hepatic venous or portal venous) was observed. In addition, there was a significant difference in microscopic portal vein invasion (p = 0.0034), microscopic hepatic venous invasion (p = 0.0130), and microscopic vascular invasion (portal or venous) (p = 0.0007) between peri-tumoral hypointensity-positive HCCs and peri-tumoral hypointensity-negative HCCs. On the other hand, there was no significant difference in microscopic portal venous invasion (p = 0.4662) and microvascular invasion (hepatic venous or portal venous) (p = 0.1828) between P-hyperintensity-positive (grade 1–3) and P-hyperintensity-negative group (grade 0). Therefore, we think that peri-tumoral hypointensity on HB phase is better imaging finding for prediction of microvascular invasion compared with P-hyperintensity.

The limitations of this study included the following. First, the histological definition of P-hyperplasia was not completely clear as mentioned above. Second, the entire circumference of the tumor was not evaluated in the histological examination, and therefore the histological analysis of P-hyperplasia was not robust. Third, the presence of artifacts and the limited resolution of the MR imaging may have modified to some degree the results of the image analyses.

In conclusion, peri-tumoral hyperintensity (P-hyperintensity) on HB phase of EOB-MRI in HCC may indicate peri-tumoral hyperplasia (P-hyperplasia) with GS and OATP1B3 expression and a higher incidence of microscopic hepatic venous invasion.