A Clinical–Radiomics Multimodal Fusion Model Integrating Preoperative CT and Intraoperative DSA for Predicting Short-Term Response to Initial TACE in Hepatocellular Carcinoma

Objective We developed a clinical-radiomics multimodal model for the prediction of the short-term efficacy of initial transarterial chemoembolization (TACE) in patients with hepatocellular carcinoma (HCC). Methods In this retrospective cohort study of 100 patients with intermediate to advanced HCC, a total of 1,037 features were extracted from contrast-enhanced computed tomography (CECT) and digital subtraction angiography (DSA) images. Minimum redundancy maximum relevance and least absolute shrinkage and selection operator (LASSO) regression were applied for feature selection and model construction. Multivariate logistic regression (LR) was used to build a clinical-imaging model based on clinical factors and a clinical-radiomics model. Model performance was evaluated using the area under the receiver operating characteristic curve (AUC). Results Among the single-modality models, the DSA-based model achieved the highest discriminative ability (AUC = 0.742), outperforming the CT-based model (AUC = 0.729) and the clinical model (AUC = 0.542). The fusion model integrating DSA_Rad_Score, CT_Rad_Score, and clinical variables demonstrated the best overall performance (AUC = 0.893). Conclusion The multimodal predictive model integrating CT and DSA radiomics features with clinical variables demonstrated superior performance in predicting short-term treatment response to initial TACE in patients with HCC.