Development and validation of an integrated machine learning model for recurrence-free survival prediction in clear cell renal cell carcinoma

Abstract

Highlights • Integrated ML framework fuses clinical features and DL-sign for ccRCC RFS prediction.• 3D-ViT outperforms ResNet variants, delivering high external test AUC of 0.846 for DL signature.• Combined nomogram achieves excellent external C-index of 0.910 with robust multi-year RFS prediction.• Model outperforms UISS/SSIGN, enabling precise ccRCC risk stratification and clinical decision support.