Predictive Radiomics-Based Model for Recurrence-Free Survival After Curative Resection in Patients with Hepatocellular Carcinoma.

Abstract

Background: Postoperative recurrence after curative resection is a major concern in the management of hepatocellular carcinoma (HCC). This study aimed to develop a radiomics-based model for predicting recurrence-free survival (RFS) after curative resection.

Methods: We retrospectively included 184 patients with early-stage HCC who underwent curative resection. The patients were randomized into training and validation sets in a 7:3 ratio. Radiomics features of the tumors on CT images were extracted to construct the Rad-score. We incorporated the Rad-score, clinical characteristics and biochemical parameters into univariate and multivariate analyses to construct a COX proportional hazards model. A radiomics-based nomogram model for predicting recurrence risk was developed by integrating multiple factors that affect recurrence. Calibration curve was used to assess the predictive performance of the model.

Results: Rad-score was constructed using 15 radiomic features. The results of multivariate analyses showed that Rad-score, lactate dehydrogenase (LDH) and alpha-fetoprotein (AFP) were independent predictors of RFS. They categorized patients into different recurrence risk groups, and RFS was significantly prolonged in patients in the low-risk group in the training (p<0.001) and validation sets (p<0.001). The Rad-score based composite prediction model showed good predictive performance with AUC of 0.765 and 0.920 for predicting 3 years RFS in the training and validation sets, respectively. The calibration curves indicated that the nomogram model had a favorable predictive performance.

Conclusion: This postoperative predictive model allows for better screening of patients at a high risk of recurrence and is a valuable instrument to guide clinicians in clinical treatment decisions.