AUTHOR=Quan Xin , Huang Xinqiao , Liu Jiong , Yuan Xiang , Shu Jian 

TITLE=Preoperative assessment of longitudinal extent in hilar cholangiocarcinoma using noninvasive enhanced MR radiomics: a multicenter study

JOURNAL=Frontiers in Oncology

VOLUME=Volume 15 - 2025

YEAR=2025

URL=https://www.frontiersin.org/journals/oncology/articles/10.3389/fonc.2025.1632630

DOI=10.3389/fonc.2025.1632630

ISSN=2234-943X

ABSTRACT=ObjectiveThis study aims to develop a noninvasive radiomics model based on magnetic resonance imaging (MRI) for accurately predicting the longitudinal extent of hilar cholangiocarcinoma (HCCA), to assist in subsequent surgical decision making.MethodsThis study retrospectively collected and analyzed data from patients with HCCA across three medical centers in China. Radiomics quantitative features were extracted from T2-weighted imaging (T2WI), diffusion-weighted imaging (DWI), and enhanced T1 high-resolution isotropic volume examination (e-THRIVE) sequences. L1 regularization was employed to select features, and three single-sequence radiomics models were developed to predict Bismuth type IV of HCCA. To improve the predictive accuracy for Bismuth type IV, the fusion model integrating the three single-sequence models was constructed. The performance of these models was evaluated comprehensively, and the optimal radiomics model for predicting longitudinal extent was identified.ResultsA total of 154 patients with HCCA were included in the analysis. The radiomics models based on T2WI, DWI, and e-THRIVE sequences demonstrated predictive capabilities, with AUC values in the training set of 0.867, 0.923, and 0.872, respectively, and AUC values in the test set of 0.809, 0.823, and 0.808, respectively. The fusion model, which combined features from all three sequences, achieved superior predictive performance, with an AUC of 0.980 in the training set and 0.907 in the test set. This model demonstrated robust potential for predicting whether the HCCA was classified as Bismuth type IV.ConclusionThe multi-sequence MRI-based radiomics model can effectively predict Bismuth type IV of HCCA, assisting in clinical surgical decision-making, facilitating R0 resection to improve the prognosis of patients with HCCA.