Development of a radiomics nomogram based on the MRI features to predict the survival of hepatocellular carcinoma patients

Abstract

Introduction: Postoperative recurrence for HCC is high, with 5-year rates reaching 70%, suggesting that even in the same early-stage, patients have a diverse postoperative prognosis. Thus, the current staging systems still need improvement. We aim to determine the integrative value of Magnetic Resonance Imaging (MRI) radiomic features, clinical, pathological, molecular and biochemical cellular markers in survival prediction for Hepatocellular Carcinoma (HCC) patients. Method: A total of 188 patients with HCC who had undergone radical resection were enrolled in this retrospective study (training cohort, n = 141; validation cohort, n = 47). Radiomic features (n=676) were extracted from pre-treatment MRI. A R-signature (Rad-score) was generated using the least absolute shrinkage and selection operator (lasso) Cox regression model. The association between the R-signature and overall survival (OS) was assessed in the training cohort and verified in the validation cohort. A radiomics nomogram was then established integrating significant clinicopathological and molecular parameters and was evaluated in both cohorts. Result: The six-feature-combined R-signature successfully stratified patients with HCC who had undergone radical resection into two prognostic risk groups in both cohorts. The radiomics nomogram incorporating R-signature, significant clinicopathological (inflow occlusion, tumor size, histological grade, microvascular invasion) and molecular parameters (GRK6, PTP4A3) estimated OS better than the clinicopathological and molecular nomogram (C-index 0.85 vs. 0.74 and 0.82 vs. 0.69, for training and validation cohorts, respectively) or R-signature alone (C-index 0.85 vs.0.76 and 0.82 vs. 0.75 for training and validation cohorts, respectively). Conclusion: The R-signature could be used to stratify patients with HCC following radical resection into high- and low-risk groups. The integrative nomogram incorporated MRI radiomic features, clinicopathological parameters and molecular and biochemical cellular markers features exhibit a favorable prognostic predictive accuracy.