CT-Based Radiomics for the Recurrence Prediction of Hepatocellular Carcinoma After Surgical Resection

PURPOSE: To explore the effectiveness of radiomics signature in predicting the recurrence of hepatocellular carcinoma (HCC) and the benefit of postoperative adjuvant transcatheter arterial chemoembolization (PA-TACE). PATIENTS AND METHODS: In this multicenter retrospective study, 364 consecutive patients with multi-phase computed tomography (CT) images were included. Recurrence-related radiomics features of intra- and peritumoral regions were extracted from the pre-contrast, arterial and portal venous phase, respectively. The radiomics model was established in the training cohort (n = 187) using random survival forests analysis to output prediction probability as “Rad-score” and validated by the internal (n = 92) and external validation cohorts (n = 85). Besides, the Clinical nomogram was developed by clinical-radiologic-pathologic characteristics, and the Combined nomogram was further constructed to evaluate the added value of the Rad-score for individualized recurrence-free survival (RFS) prediction, which is our primary and only endpoint. The performance of the three models was assessed by the concordance index (C-index). Furthermore, all the patients were stratified into high- and low-risk groups of recurrence by the median value of the Rad-score to analyze the benefit of PA-TACE. RESULTS: The model built using radiomics signature demonstrated favorable prediction of HCC recurrence across all datasets, with C-index of 0.892, 0.812, 0.809, separately in the training, the internal and external validation cohorts. Univariate and multivariate analysis revealed that the Rad-score was an independent prognostic factor. Significant differences were found between the high- and low-risk group in RFS prediction in all three cohorts. Further analysis showed that compared with the low-risk group, patients with the high-risk received more benefits from PA-TACE. CONCLUSION: The newly developed Rad-score was not only a powerful biomarker in predicting the RFS of HCC but also a strong stratification basis to explore the high-risk patients who could benefit from PA-TACE.