Imaging biomarkers for well and moderate hepatocellular carcinoma: preoperative magnetic resonance image and histopathological correlation

Hepatocellular carcinoma (HCC) is one of the most common malignancies and is associated with poor prognosis [1]. Despite the development of various treatments, including surgical resection, liver transplantation, transcatheter arterial chemoembolization (TACE), and radiofrequency ablation, postoperative recurrence remains the most important prognostic factor [2]. Some studies have reported that tumor size, number of lesions, presence of vascular invasion, and degree of tumor differentiation are important factors influencing the early recurrence of tumors [3‐6]. Edmondson-Steiner grade might be an independent factor affecting prognosis or recurrence of HCC. Cucchetti, et al. [7] reported that the postoperative recurrence rate of high histological grade HCC was twice as high when compared to low grade HCC. The structure of intratumoral vessels is also related to the histological grade of HCC [8, 9].

On the other hand, preoperative histological grading of HCC is a main parameter in planning of therapeutic approach. For example, in surgical resection, the internationally recognized tumor-free margin is 1 cm around the lesion [10]. However, if the preoperative pathological grade is high, the resection range can be extended under the condition that normal liver function can be maintained post resection. According to some research, when the tumor-free margin is extended to 2 cm, the recurrence rate of HCC decreases markedly [11]. Furthermore, for postoperatively diagnosed high-grade HCCs, the follow-up period should be shortened and the frequency should be increased. More accurate inspection methods, such as dynamically enhanced magnetic resonance imaging (MRI), may also be applied to detect early recurrence or metastasis. Consequently, the role of pre-operative imaging for the assessment of well, moderate and poor differential HCC is crucial.

Preoperative pathological grading of HCC mainly relies on biopsy. However, due to the typical heterogeneity of the lesion, focal biopsy does not necessarily reflect the overall state of the tumor [12‐14]. Furthermore, needle biopsy is invasive and is associated with complications such as bleeding and tumor implantation [15, 16]. Imaging methods can give an overall perspective on the tumor, which is of great significance in the preoperative evaluation of HCC. MRI is one of the most important imaging methods for liver diseases, owing to its good soft tissue resolution and multiple sequences and parameters. The application of liver-specific contrast agents has further improved the value of MRI in liver diagnosis. Gadoxetic acid is a widely applied liver-specific contrast agent. Hepatobiliary phase images obtained after gadoxetic acid enhancement not only improve lesion detection and diagnosis, but also reflect the biological behavior of a tumor at the molecular level [17, 18].

Imaging findings for HCC are closely related to the pathology of the lesions. Some studies have found that some features of contrast-enhanced ultrasound and computed tomographic (CT) images were associated with higher pathological grade of HCC [19, 20]. Heo et al. reported that the apparent diffusion coefficient (ADC) values of poorly differentiated HCCs were lower than those of well-differentiated HCCs [21]. Other studies have found that MRI enhancement patterns were closely related to the pathological grading of HCC [21‐23]. However, most previous studies have focused on only one imaging characteristic, such as the tumor signal or enhancement pattern, to evaluate HCC histological grade. Few studies have comprehensively assessed the relationship between multiple imaging features and HCC histological grading [22, 24].

The purpose of this study was to analyze various qualitative and quantitative imaging features of gadoxetic acid-enhanced MRI in relation to the histological grading of single HCC lesions, and to identify the MRI features associated with high grade HCCs, in order to provide guidance for clinical treatment.

This study received approval from the Institutional Review Board (IRB). Written informed consent was obtained from each patient.

This study found that the tumor maximum diameter was an independent predictor of HCC histological grading. Other imaging features, including irregular tumor margin, presence of intratumoral vessels and peritumoral hypointensity were also associated with the higher histological grading of HCC. Other qualitative MRI features such as tumor thrombus, signal intensity, tumor capsule, tumor margins, and peritumoral enhancement, as well as quantitative features including ADC value and percentage decrease in T1 (T1_D%), had no statistical correlation with HCC histological grades.

Tumor maximum diameter is an important prognostic factor for HCC. With each increase in tumor size, HCCs become more prone to blood vessel invasion and extrahepatic metastasis, which are associated with higher histological grades [27, 28]. According to Lee et al., 86% of medium-grade HCCs are more than 20 mm in diameter, compared to 52% of low-grade HCCs [24]. Our results also demonstrated that the maximum diameters of tumors in medium-grade HCC were larger than those in low-grade HCC. A total of 79% of the HCCs with maximum diameter less than 53 mm were low-grade HCCs, while 51% of HCCs with diameters more than 53 mm were medium-grade. Similarly, Pawlik et al. also found that the HCCs with diameter greater than 50 mm had higher histological grades [29].

The univariate analysis showed that imaging features including tumor margin, intratumoral vessels, and peritumoral hypointensity were related to HCC histological grading. However, the multivariate analysis found that these characteristics are not independent risk factors for HCC. This outcome may be a result of interaction among these features in the multivariate analysis. According to the classification method of the Japanese liver cancer research team, the edge of the tumor can be categorized into five forms [30]. As did Chou’s study [31], our study included three of these categories, the single nodule, focal protruding nodule, and multinodular fusion types. We found that multinodular fusion HCC tends to have higher histological grade compared with focal protruding nodular and single-nodular HCC. Another study found that most HCCs with irregular margins were moderately or poorly differentiated, and that the sensitivity and specificity were 82.4 and 47.6%, respectively [24]. On the other hand, a different study reported that there was no association between the tumor margin and the histological grade of HCC [22]. However, in that study, tumor margin evaluation was performed on conventional gadoxetic acid-enhanced images. Therefore, the results may have been biased by factors such as peritumoral enhancement. In comparison, gadoxetic acid is a hepatobiliary-specific MRI contrast agent. After intravenous injection, most HCCs is characterized as an area of relatively low signal without gadoxetic acid uptake during hepatobiliary phase, while the surrounding normal liver parenchyma present high signal because of their high uptake ability for gadoxetic acid. Hence, the difference in signal between HCC and surrounding liver parenchyma becomes more significantly during the hepatobiliary phase of gadoxetic acid enhanced MRI than conventional contrast agents. This makes the boundaries of tumors clearer to delineate where a more acurate evaluation of tumor margin becomes possible [32].

Presence of intratumoral vessels are an important feature for the diagnosis of HCC. The positive predictive value for HCC is 90% when a liver lesion demonstrates visible intratumoral vessels, with a specificity of 98% [33]. According to Lee et al. [24], about 65% of high-grade HCCs were found to contain intratumoral vessels, and about 64% of medium-grade HCCs were associated with intratumoral vessels; however, intratumoral vessels were absent in 71% of low-grade HCCs [24]. Consistent with the results of Lee’s [24] study, our results showed that HCCs with intratumoral vessels tended to have higher histological grades than those without intratumoral vessels. In our study, 70% of medium-grade HCCs were found to contain intratumoral vessels, but only 38% of low-grade HCCs.

Peritumoral hypointensity refers to an area around the tumor, which is hyperintense to the tumor tissue but hypointense to the normal hepatic parenchyma during the hepatobiliary phase [34]. In this study, we found that 48% of medium-grade HCCs showed hypointensity around the tumor, but only 18% of low-grade HCCs. According to previous reports, occlusion of micro-portal veins caused by microvascular tumor thrombus and insufficiency of the arterial blood supply result in the dysfunction of local liver cells, which subsequently leads to decreases in the uptake of gadoxetic acid, resulting hypointensity around the tumor during the hepatobiliary phase [35‐37]. Therefore, we speculate that the peritumoral hypointensity of HCC with high histological grade is associated with microvascular invasion.

Many studies have shown that vascular invasion occurs often in high-grade HCC [38‐41]. Vascular invasion can be divided into microvascular invasion and macrovascular invasion [42]. Currently, microvascular invasion cannot be directly visualized on imaging [43]. The venous tumor thrombus in this study only involved large veins, including large branches of the portal vein and hepatic vein. Although many previous studies have shown that vascular invasion is closely related to the level of tumor tissue differentiation [38‐41], this study did not find a relationship between venous tumor thrombi visible on MRI and the histological grade of HCC. This may be a result of selection bias in the study sample, and a larger sample size is required for further validation.

Witjes et al. [22] reported no differences in the incidence of heterogeneity on T2WI among different levels of HCC differentiation. Our study also found no correlation. However, Lee et al. [24] found that HCC with heterogeneity tended to have poor differentiation and high histological grade, and that medium- or high-grade HCCs were more susceptible to necrosis, resulting in heterogenous density on CT images.

In this study, equilibrium phase images were used to evaluate the existence and integrity of a tumor capsule, which is defined as a hyperintense ring around the lesion in delayed or equilibrium phase contrast enhanced MRI [44]. The delayed enhancement of the tumor capsule is associated with slower blood flow in micro-vessels [44, 45]. However, not all HCCs demonstrate tumor capsules: previous studies suggest 24–90% of Asian HCC patients, and 12–42% of non-Asian HCC patients demonstrates tumor capsules [46‐50]. Some researchers have suggested that a tumor capsule is one of the characteristics of high-grade HCC [48, 49]. In this study, there was no correlation between the tumor capsule status on contrast-enhanced MRI and the histological grade, which is consistent with the results of Lee’s [24] study. Possible explanations are that the degree of HCC differentiation has relatively little effect on the tumor capsule, or that the tumor capsule of some lesions is very thin and may not be observed on MRI.

Although the mechanisms of peritumoral enhancement are not fully understood, most researchers believe that it may be caused by microvascular invasion, which causes occlusion of the microportal veins and a compensatory increase in hepatic arterial blood supply [51]. Studies have shown that high-grade HCC is prone to microvascular invasion [52, 53]. In this study, however, no correlation was found between peritumoral enhancement and HCC histological grade. This may be related to the limited sensitivity of MRI to peritumoral enhancement.

ADC value had been use to assess grade of HCC in previous studies have. Heo et al. [21] found that the ADC value and the degree of HCC differentiation were negatively correlated. Nakanishi et al. [54] reported that the ADC value in high-grade HCC was significantly lower than in medium- and low-grade HCC [55]. However, the results of our study demonstrated no statistical relationship between HCC grade and ADC values. The difference may be due to the relative larger lesions in our research which represent higher degree of degeneration and necrosis. At the same time, not all the patients in the study were received DWI. Therefore, selective bias could not be fully avoided. On the other hand, the instability of ADC value might be another limitating factor. The determination of the ADC value is easily influenced by many factors, such as differences in the heartbeat and respiratory control of patients, due to the intrinsical sensitivity of echo echoplanar imaging (EPI) sequence to motion artifacts [55].

Peng et al. [56] showed that the percentage of T1 reduction was different before and after gadoxetic acid-enhancement in HCCs with different pathological grades. However, our results showed no correlation between the percentage of T1 reduction and the histological grade of HCC. This difference may be a result of selection bias in the samples. The study of Peng included 53 HCC lesions, all of which included T1 mapping images, and among which 13 lesions were pathologically diagnosed as grade I. In this study, only 97 cases included T1 mapping images, and only 8 cases were diagnosed as grade I HCC. In this study, we combined grade I and grade II HCC together as low-grade group, and grade III as medium grade group, to evaluate the difference of T1_D% between low and medium grade group. Because there were only 8 cases of grade I HCC, the result of this study is also equivalent to a finding that the T1_D% of grades II and III showed no statistically significant difference, which is similar to the results of Peng’s study. However, the study sample was relatively small and further study is needed to confirm the role of T1 mapping for prediction of HCC grading.

In this study, there were several imaging features (tumor margin, intratumoral vessels, peritumoral hypointensity, and maximum tumor diameter) which demonstrated statistically significant correlation with the histological grade of HCC during univariate logistic analysis. This will be due to the failure of adjustment the interaction effect of the factors for univariate logistic analysis. When many factors were excluded from each other in multivariate logistic regression analysis, only the maximum tumor diameter was an independent risk factor for high grade HCC. Tumor diameter has been included in liver cancer staging systems, such as the TNM staging system and Milan liver transplantation criteria [57, 58]. Consequently, tumor diameter has been showed as the most significant influence on the biological behaviors, such as grading, microvascular invasion, recurrence or prognosis in HCC.

There are some limitations in this study. First, most of the HCC lesions in this study were large nodules or masses, the results of this study may not be applicable for small HCCs. Second, the cohort in this study is heterogeneous because mostly consisted of grade 2 and grade 3 HCCs, with only a few grade 1 HCCs and no grade 4 HCC.. Third, some of the patients in the study had not the quantitative MRI features (ADC value, T1 _D%). Thus, the selective bias could not be fully avoided.

In conclusion, greater size, a more irregular margin, presence of intratumoral vessels, and peritumoral hypointensity during HBP indicate a higher histological grade of HCC. The maximum tumor diameter is an independent risk factor for high grade HCC. Our study has demonstrated that the radiological prediction of HCC grades is possible before treatment is practicable which can help determine the most appropriate surgical strategies for patients based on more accurate risk-benefit assessment. .