Explainable application of AI machine learning in the individualized survival prediction of node-pos

An international medical study has yielded new findings with relevant findings that are of important reference value for disease prevention codes. Complex predictions of regional lymph nodes involved but not distant transfer (N1M0) This study aims to develop and validate an explanatory machine learning model that predicts specific cancer-specific survival (CSS) 18. 287 N1M0 patients.

18,287 N1M0 patients' data in the SEER database are divided into training (n = 4780), internal testing (n = 1193), and two temporary validation cohort (n = 3149). n = 9165);; Cox proportional hazards and four machine learning models were compared using the C-index, time-dependent AUC, and Integrated Brier Score;;SHAP was used for interpretation, while IPTW was used for sensitivity analysis;; Random Survival Forest (RSF) has a higher C-index of 0. 697 internal; 0.

740 temporal validation in all models. SHAP ((Survival Heterogeneity Analysis) reveals that deep prostate removal is the strongest protective factor, followed by lower T-cell and PSA levels. IPTW ((International Prostate Cancer Foundation) confirms positive local control survival benefits.

The RSF model provides accurate, robust, and explainable background information that is important as a guide for patients with N1M0 atypical prostate cancer. As a web-based calculator, it delivers accurate risk categorization and personalized treatment planning.