Hepatocellular carcinoma (HCC) is the most common primary liver cancer, frequently occurring in patients with chronic liver disease or cirrhosis. HCC is characterized as a heterogeneous tumor considering its genetic and pathologic features and prognosis. The well-known prognostic factors of HCC include serum markers such as serum levels of α-fetoprotein (AFP) and protein induced by vitamin K absence or antagonists-II (PIVKA-II), pathologic features such as differentiation, microvascular invasion, satellitosis, subtype, immunohistochemical expression of keratin 19 (K19), and genetic features such as fibroblast growth factor 19 amplification and proliferative class.1 Imaging findings representing the extent of tumor, including tumor size, number, gross vascular invasion, and extrahepatic metastasis, have also been recognized as prognostic factors and incorporated into current staging systems.2-4

Cross-sectional imaging findings play a crucial role in the diagnosis of HCC. Liver magnetic resonance imaging (MRI) protocols commonly consist of T2-weighted images, dual gradient-echo images to assess the presence of fat and iron, fat-suppressed T1-weighted dynamic enhancement images to characterize dynamic enhancement patterns, and diffusion-weighted images to assess the free diffusion of water molecules, which are known to be related to tumor cellularity or necrosis. Hepatobiliary phase images are added when gadoxetic acid (Gd-EOB-DTPA, Bayer HealthCare Pharmaceuticals, Berlin, Germany) is used as the contrast medium. Particularly, imaging findings in dynamic phases employing intravenous contrast agents are most important for HCC imaging diagnosis, since they reveal characteristic vascular changes during multistep hepatocarcinogenesis. In addition, the hepatobiliary phase is another key sequence that shows the difference between HCC and non-tumor livers. The uptake of gadoxetic acid by normal hepatocytes occurs via organic anion transporter polypeptide 8 (OATP8); the period when this uptake is most evident is called the hepatobiliary phase, which corresponds to approximately 15–20 minutes after intravenous injection of gadoxetic acid. Since the expression of OATP8 and the hepatobiliary uptake of gadoxetic acid decline progressively during multistep hepatocarcinogenesis, HCC exhibits a lower hepatobiliary uptake and signal intensity than non-tumor liver.5

The contemporary guidelines of HCC recommend imaging-only diagnosis of HCC without pathological confirmation when a hepatic lesion shows typical imaging findings of HCC in individuals at a high risk of HCC.6-8 Since several HCC cases are diagnosed based on imaging findings, without obtaining tissue, it is difficult to apply genetic and pathologic prognostic factors in such cases. Recent studies have shown that imaging findings of HCC are not only relevant for the diagnosis of HCC, but also for identifying genetic and pathologic characteristics and determining the prognosis (Table 1).9-12

Here, we review the imaging features of HCC associated with the prognosis of HCC patients.

Fat in mass is one of the Liver Imaging Reporting and Data System (LI-RADS) ancillary features favoring HCC in particular.13 The loss of signal intensity on T1-weighted out-ofphase images compared with in-phase images of dual gradient-echo sequences suggests the presence of fat.14 Because fat is rarely found in pure cholangiocarcinoma, the presence of intralesional fat is useful for the differential diagnosis between hepatocellular lesions and cholangiocarcinoma.15 The mechanism by which fat develops inside HCC is presumed to be clonal expansion of hepatocytes with anomalous fat metabolism and cellular metabolic disturbances due to switching of the dominant blood supply from portal venous to hepatic arterial and the consequent ischemic/hypoxic conditions.16,17 It is most often observed in early HCC <1.5 cm in diameter and tends to decrease with increasing tumor size and grade (Fig. 1).17 However, in steatohepatitic HCC, which is most commonly found in background steatohepatitis or non-alcoholic fatty liver disease, intralesional fat can be observed not only in early HCC but also in advanced tumor grades.18-20

It has been reported that radiological or histological fat in mass is associated with infrequent microvascular invasion (MVI), suggesting that fat in mass may be associated with a better prognosis.18,21,22 Indeed, several studies have reported an association between fat in mass and a favorable prognosis after curative treatment. Siripongsakun et al.23 reported fewer distant metastases (4.3% vs. 21.7%) and longer time to progression in fat-containing HCCs than in non-fat-containing HCCs in a case-control study of patients who received local treatment or liver transplantation. A study conducted by Chen et al.24 showed that fat in mass was a favorable prognostic factor for tumor recurrence in patients with LR-5 HCCs who underwent hepatic resection. Another study suggested that fat in mass was a favorable prognostic factor for patients undergoing radiofrequency ablation.25

During multistep hepatocarcinogenesis, HCC becomes hypervascular with the development of unpaired arteries and sinusoidal capillarization.26 Arterial phase hyperenhancement (APHE) is one of the most important imaging findings in HCC diagnosis. Typically, HCC shows APHE of non-rim pattern; LI-RADS defines it as “non-rim like enhancement in arterial phase unequivocally greater in whole or in part than liver.”13

The pathological fibrotic capsule is considered to be the result of compression of peritumoral fibrous tissue by the tumor and is closely related to the gross morphology of HCC.48,49 HCCs showing a vaguely nodular type, mostly early HCCs, do not show a fibrotic capsule because of lepidic growth. Fibrotic capsules are mainly observed in nodular HCCs, including HCCs of the single nodular type, single nodular with extranodular growth type, and multinodular confluent type. The fibrotic capsule temporarily serves as a physical barrier to block tumor infiltration; however, capsular and extracapsular tumor infiltration emerge as HCC progresses, and eventually the capsule is seldom observed in the infiltrative type of HCC.48,50 Collectively, the pathological fibrous capsule is not commonly observed; in early HCC it shows vaguely nodular margins and in advanced HCC it shows infiltrative margins.

“Enhancing capsule” is one of the major imaging features for diagnosis of HCC in LI-RADS. It refers to the observation of a smooth, uniform, sharp border around most or all of an observation in the portal venous phase, delayed phase, or transitional phase (Fig. 7).13 The fibrous capsule is strongly associated with enhancing capsule appearance although this is not always the case. Approximately 14–17% of the enhancing capsules are not pathologically fibrous capsules, but pseudo-capsules consisting of prominent sinusoids and peritumoral fibrosis mimicking bridging fibrosis.51,52

Some studies have indicated that the presence of an enhancing capsule on imaging may be a good prognostic factor in patients undergoing hepatic resection or transarterial chemoembolization.53-55 However, other studies have shown that the enhanced capsule is not significant for prognosis or related to a worse prognosis.27,56 The inconsistent results regarding the prognostic significance of the pathologic or radiologic capsule in literature may be attributed to the heterogeneity of the lesions included in each study.

In multistep hepatocarcinogenesis, OATP8 expression gradually decreases, resulting in a decrease in hepatobiliary phase signal intensity on MRI, and these changes are known to precede the formation of unpaired arteries.57 Therefore, hepatobiliary phase hypointensity is a useful imaging finding for malignant transformation of hepatic nodules. In contrast, hepatobiliary phase isointensity represents the normal functioning of hepatocytes and biliary drainage inside the lesion, which generally suggests benignity.13

A low apparent diffusion coefficient (ADC) is a common radiological finding observed in various malignant tumors. Theoretically, as cell density increases and the nucleus/cytoplasm ratio increases in neoplastic tissue, it results in decreased free diffusion of water molecules, which can lead to increased signal intensity in diffusion-weighted imaging (DWI) and a decrease in the ADC value (Fig. 3).77 Both the increased DWI signal and decreased ADC value suggest restricted diffusion of water molecules, but the DWI signal intensity also reflects the T2 signal. In contrast, as ADC excludes the effect of the T2 signal from DWI, it is regarded as a reliable index that more accurately depicts diffusion restriction. Contrarily, intratumoral necrosis can decrease signal intensity in DWI and increase ADC values.78

Low ADC values have been reported to be related to poorer histological grade,79-81 presence of MVI,82-84 expression of progenitor cell markers,85,86 and proliferative signatures of HCC.87-89 A few studies have reported that a low ADC value is a significant predictor of early recurrence after curative resection of HCC.79,90

ADC values are quantitative, but ADC values and diffusion-weighted signal intensities vary substantially depending on the imaging technique and MRI scanner; therefore, it is difficult to suggest a generalizable cutoff value that can be used clinically.91 Additionally, HCC is frequently accompanied by necrosis, and the ADC value may vary significantly depending on the location of the measurement within the lesion.

LR-M is a diagnostic category of LI-RADS for malignant tumors, not definitely HCC. Tumors of LR-M include targetoid masses or non-targetoid masses with the following appearances: infiltrative appearance, marked diffusion restriction, necrosis or severe ischemia, and other features suggestive of non-HCC malignancy.13 A targetoid appearance (concentric arrangement of internal components), reflects peripheral hypercellularity, central stromal fibrosis, or ischemia, which are observed in various phases or sequences. This includes peripheral washout, delayed central enhancement (Fig. 4), targetoid diffusion restriction (Fig. 3), and targetoid appearance in the transitional or hepatobiliary phase, in addition to the rim-APHE described above. Although it is frequently reported in non-HCC malignancies, such as intrahepatic cholangiocarcinoma, combined hepatocellular cholangiocarcinoma, or metastasis, 22–36% of LR-M lesions were found to be HCC, because of the high prevalence of HCC in patients with underlying chronic liver disease.92-95

HCCs showing LR-M features have been reported to be associated with MVI,67 poor histological differentiation.96 and stem/progenitor marker expression, including K19 and EpCAM.76,97 LR-M HCCs are likely to have a higher risk of early recurrence and poor overall survival after resection and a higher risk of tumor recurrence five years after liver transplantation.36,67,96,98 A study by Choi et al.98 demonstrated such trends in all primary liver cancers, as well as HCCs with LR-M features. Recently, Moon et al.99 reported that the presence of rim-APHE was an independent prognostic factor for postoperative survival, while the 5-year overall survival and recurrence-free survival of LR-M HCCs without rimAPHE were not different from those of LR-4/5 HCCs. However, whether rim-APHE is a more important prognostic factor than the other LR-M features needs to be validated in other studies.

Several imaging findings that indicate a poor prognosis, such as rim-APHE and non-smooth tumor margins, arterial phase peritumoral enhancement, and peritumoral hypointensity in the hepatobiliary phase, usually appear in combination. Currently, there is no consensus regarding which of these imaging findings is the most significant prognostic factor. One of the reasons for this variability may be that different studies use different definitions for imaging findings, and a more standardized definition of each imaging finding is needed to ensure consistent evaluation and validation of their prognostic significance. Moreover, the substantial interobserver variability of these imaging results has been identified as a drawback. Min et al.100 reported that there was considerable interobserver variability in each radiologic finding or combination (κ=0.38–0.47) and predicted the probability of MVI based on imaging findings (κ=0.41) evaluated on gadoxetate-enhanced MRI. For HCCs <3 cm, there was no difference in interobserver agreement for MVI according to the reviewers’ experience (κ=0.43 vs. 0.47, P>0.999), while for HCCs >3 cm, more experienced reviewers showed higher agreement than the less experienced reviewers (κ=0.65 vs. 0.21, P<0.001). Quantitative imaging analysis such as radiomics, can be utilized to reduce interobserver variability; however, the results of quantitative analysis vary significantly with imaging instruments and protocols of imaging studies.

The majority of research has been conducted in Northeast Asia, where hepatitis B is the most prevalent cause of liver disease, and the prognosis of patients with favorable liver function who underwent hepatic resection has been investigated. In addition, most studies to date have been conducted retrospectively at a single institution with relatively small sample sizes. Therefore, it is not clear whether it can be generalized and applied to Western patients or to advanced HCCs not undergoing hepatic resection. Currently, these predictive imaging results are not incorporated into major HCC treatment guidelines because there are no well-validated imaging findings. A large-scale, multicenter study is needed in the future.

Recent studies have demonstrated the possibility that imaging features can predict not only pathological findings, such as histologic grade, presence of MVI, and pathological subtypes of HCC, but also the risk of tumor recurrence and survival. The imaging phenotype consisting of each radiological feature and/or its combination has the potential to be used for individualized treatment decisions in the future.