Performance of multi-regional radiomics features and clinical-radiological variables in the prognostic analysis and risk stratification of single hepatocellular carcinoma

Abstract

Abstract Background To develop multi-regional radiomics models to evaluate the prognosis of single hepatocellular carcinoma (HCC) after hepatectomy and stratify risk by combining radiomics features with clinical-radiological variables. Methods This retrospective study enrolled 207 patients with single HCC after surgery (training set:validation set = 144:63). Different volumes of interest (VOIs) were constructed to extract radiomics features, and the model with the largest area under the receiver operating characteristic curve (AUC) was considered optimal. Prognostic clinical-radiological variables were identified via univariate and multivariate Cox regression analyses. A combined model incorporating radiomics features and clinical-radiological variables was utilized to predict outcomes and stratify recurrence risk. The Kaplan–Meier method and the log–rank tests were applied to estimate recurrence-free survival (RFS). Calibration curves and decision curve analysis were employed to assess performance of the combined model. Results Among the multi-regional radiomics models, the model based on VOItumor + 5mm had the highest AUC of 0.803. Multivariate analysis identified age, cirrhosis, hepatitis, albumin-bilirubin grade, gamma-glutamyl transpeptidase level, microvascular invasion, and mosaic architecture as risk factors of RFS. The combined model achieved C-indexes of 0.820 and 0.694 in the training and validation sets, respectively. The Kaplan–Meier curve, calibration curve, and decision curve analyses suggested that the combined model might be a non-invasive and reliable approach for assessing clinical outcomes and accomplishing relapse risk stratification. Conclusion The VOItumor+5mm-based radiomics model in conjunction with clinical-radiological variables can satisfactorily stratify risk for patients with single HCC ≤ 5 cm.