Motivation: Mutations in the β-catenin gene are significantly associated with immune evasion and prognosis in HCC patients. Goal(s): Develop and validate radiomics models using dynamic contrast-enhanced MRI to identify β-catenin status and prognosis in HCC. Approach: 465 HCC patients from four centers were enrolled. Tumor boundaries were delineated, and six machine learning algorithms were used to predict β-catenin mutation status. Ten radiomics models, a clinical model, and combined clinical-radiomics models were developed. Results: The GBM-based radiomics model outperformed others. The ADC model was the best single-sequence, while the all-sequence model was the most effective combined model. Integrating clinical factors achieved the highest predictive performance. Impact: The radiomics model using DCE-MRI and clinical factors offers a new tool for personalized treatment.
Chen et al. (Tue,) studied this question.