MRI-Derived Intra- and Peritumoral Heterogeneity for Predicting Response to Transarterial Chemoembolization in Patients with Hepatocellular Carcinoma.

[RATIONALE AND OBJECTIVES] The compromised therapeutic efficacy of transarterial chemoembolization (TACE) in hepatocellular carcinoma (HCC) is closely associated with tumor heterogeneity. Magnetic resonance imaging (MRI)-derived quantification of intratumoral and peritumoral heterogeneity (ITH and PTH) may provide novel, noninvasive biomarkers. To develop and validate an MRI-based strategy leveraging quantitative ITH and PTH for pretreatment identification of patients with HCC achieving objective response to TACE monotherapy.

[MATERIALS AND METHODS] This retrospective study included patients with intermediate-stage HCC who underwent TACE. Tumoral and peritumoral regions on pretreatment MRI were partitioned into distinct imaging habitats using unsupervised clustering. High-dimensional radiomic features were extracted from each habitat, and inter-habitat feature dispersion was quantified to yield ITH and PTH metrics. Performance for discriminating responders from non-responders was evaluated using receiver operating characteristic (ROC) analysis, calibration curves, decision curve analysis, and confusion matrices. Shapley additive explanations were applied to interpret model decision-making.

[RESULTS] A total of 156 patients (139 men; age, 60.52 ± 9.90 years; 91 responders) were included and split into a training cohort (n = 124) and a testing cohort (n = 32). Ten ITH/PTH features were selected to construct the ITH/PTH-driven model (TH). Ascites and enhancement pattern were identified as independent predictors of TACE response and were used to build a clinical-radiological model (Cli-Rad). A combined model integrating Cli-Rad and TH features (Cli-Rad-TH) achieved the highest discrimination for predicting TACE response (area under the receiver operating characteristic curve (AUC) = 0.899-0.901), followed by TH (AUC = 0.822 and 0.884) and Cli-Rad (AUC = 0.663-0.704).

[CONCLUSION] MRI-derived quantification of ITH and PTH enables noninvasive pretreatment prediction of TACE response. This strategy may support individualized treatment stratification and improve clinical decision-making.