AI-Based Quantification of Enhancing Tumor Volume on Contrast-Enhanced MRI to Predict Pathologic Response and...

To explore the diagnostic and prognostic value of AI-quantified MRI tumor volume for assessing pathologic response in unresectable hepatocellular carcinoma (uHCC) after hepatic arterial infusion chemotherapy plus targeted therapy and immunotherapy (HAIC-TI).

This retrospective study included 35 patients (46 lesions) who underwent HAIC-TI followed by hepatectomy. AI was used to calculate the tumor enhancement volume ratio (TEVR) from MRI. Correlation analysis was conducted to evaluate the relationship between TEVR and pathological tissue proportions. Receiver operating characteristic (ROC) curve determined the optimal cutoff for the ratio of viable tumor cells (RVTCs) to define major pathological response (MPR). The diagnostic performance of AI for MPR and its prognostic significance in recurrence-free survival (RFS) were assessed.

TEVR in portal venous phase is strongly correlated with non-necrotic tissue ratio (r = 0.89, p < 0.001). RVTCs ≤ 10% predicted reduced intrahepatic recurrence (Area Under the Curve [AUC] = 0.808, p < 0.001) and independently associated with prolonged RFS (HR [hazard ratio] = 0.19, 95% CI [confidence interval]: 0.05-0.69, p = 0.011). TEVR ≤ 19.5% in the portal venous phase demonstrated high diagnostic performance for identifying MPR (AUC = 0.879) and was significantly associated with improved RFS in both univariable analysis (HR = 0.34, 95% CI: 0.12-1.00, p = 0.049) and the multivariable model incorporating only clinical and imaging factors.

AI-based MRI quantification of TEVR effectively reflected pathologic response and served as a non-invasive prognostic marker for postoperative recurrence in uHCC patients after HAIC-TI.