Our study revealed that in high-risk patients, iCCA differed from HCC in LI-RADS-based MRI features in general, and the LR-M category facilitated the accurate classification of most iCCAs. In the meantime, LI-RADS-defined risk factors tended to alter the imaging characteristics of iCCAs, leading to a relatively higher prevalence of HCC major features and increased classification as probable HCC. Notably, the LR-5 category demonstrated high specificity in excluding the diagnosis of iCCA. Additionally, among high-risk patients with solitary resected iCCA or HCC, RFS prognosis varied based on both pathologic diagnosis and LI-RADS category, the LI-RADS category showed a stronger correlation with patient outcomes than pathologic diagnosis.

Compared to HCC, iCCA exhibited distinct MRI features in high-risk patients, with features of non-peripheral washout, enhancing capsule, severe necrosis or ischemia, and blood products as independent imaging findings for differentiation, consistent with prior studies [2, 17,18,19]. The LR-M category enabled the accurate classification of most iCCAs, achieving high specificity and diagnostic accuracy. The targetoid appearance was corresponded to the LR-M category, which is an introduced imaging feature favoring iCCA, reflecting the heterogeneous distribution of cellular and fibrotic components [20]. Most iCCAs were categorized as LR-M when presenting at least one targetoid imaging feature, with the best diagnostic performance and a highest accuracy of 82.7% in our study. Cannella et al. pointed out that the targetoid appearance on T2-weighted images exhibited high specificity for non-HCC malignancies (iCCA and combined hepatocellular-cholangiocarcinoma), which could be added as a valuable diagnostic feature [15]. However, in our high-risk cohort, the T2 targetoid appearance did not significantly enhance the diagnostic performance for iCCA. This discrepancy may be explained by the differences in cohort composition, as Cannella et al. included both iCCAs and combined hepatocellular-cholangiocarcinomas, irrespective of risk factors.

Meanwhile, we noticed that a considerable proportion of iCCAs in high-risk patients exhibited imaging characteristics that overlapped with HCC diagnostic criteria, with about 26% iCCAs presenting at least one major feature and 15% miscategorized as LR-4 or 5. These findings were in line with Cheng et al.’s data, which indicated that at least 20% of iCCAs met the radiologic criteria for HCC (i.e., APHE and/or non-peripheral venous washout) [21]. The proportion may even increase in patients with LI-RADS-defined HCC risk factors [10]. Our finding that the presence of LI-RADS defined risk factors increased the likelihood of iCCAs exhibiting HCC-like imaging features raised concerns of potential misdiagnosis, which would impact therapeutic course and outcomes. However, although 15% iCCAs were miscategorized as probably or definitely HCC, only one iCCA case was categorized as LR-5. Thus, we propose that the LR-5 category is highly specific for ruling out the diagnosis of iCCA in high-risk patients. Given that iCCA is far less prevalent than HCC in high-risk populations, the overall impact on LI-RADS miscategorization for iCCA would likely be modest.

Our results further confirmed that the LI-RADS categorization could serve as a prognostic indicator for postsurgical outcomes in high-risk patients. Although RFS prognosis differed according to both pathologic diagnosis and LI-RADS category, the LI-RADS category was correlated with postsurgical RFS independent of pathologic diagnosis. These findings coincided with prior researches [14, 20], although prior researches primarily focused on patients with cirrhosis. Our study verified this standpoint in a broader high-risk population with solitary resected iCCA and HCC, beyond those with cirrhosis. It is worth noting that most studies, including ours, have emphasized the significance of targetoid LR-M features, particularly rim APHE, which is the most frequent targetoid appearance. The LR-M imaging appearance may be associated with more aggressive tumor behavior and unfavorable prognosis, regardless of pathological type, suggesting a more proactive treatment strategy in clinical practice.

There were several limitations for our study. First, being a retrospective study and the inclusion of only pathologically-proven primary liver cancers by surgery, selection bias could be introduced, our findings require validation in prospective studies with a larger sample size. Second, we aimed to explore the impact of LI-RADS-defined risk factors on the diagnosis and prognosis of iCCA, the proportions of HBV and cirrhotic patients may limit generalization to other populations. Third, as our study focused on iCCA, we enrolled 1:1 matched HCC for comparison following a prior study [14]. Therefore, the ratio in our study may not represent the actual prevalence in the general population. However, this approach created a cohort enriched for iCCA effectively. Lastly, extracellular contrast-enhanced MRI was utilized in this study, precluding an analysis of hepatobiliary-phase LI-RADS features. Nevertheless, increasing evidence has suggested the inferiority of hepatobiliary contrast agent MRI than extracellular contrast agent MRI due to suboptimal arterial phase quality and challenges in the depiction of washout and enhancing capsule on post-arterial phase images [22].

In conclusion, in patients with LI-RADS defined risk factors, although the LR-M category effectively classified most iCCAs, these tumors frequently exhibited HCC major features, leading to miscategorization as probable HCC. However, the LR-5 category remained highly specific for excluding iCCA. Furthermore, for the high-risk patients with solitary resected iCCA or HCC, the LI-RADS category enabled the prediction of postsurgical prognosis, independently from pathological diagnosis.