AI Models for Predicting Early Prostate Cancer Recurrence Post-Robotic Prostatectomy with Ultrasound Insights

This study aimed to develop and validate artificial intelligence (AI) models to predict biochemical recurrence (BR) in patients with localized prostate cancer (PC) after robot-assisted radical prostatectomy (RP). A total of 1,024 patients were included, with 476 experiencing recurrence and 548 without. Various AI algorithms, including Artificial Neural Networks (ANN), Support Vector Machines (SVM), Random Forests (RF), Extreme Gradient Boosting (XGBoost), and Deep Neural Networks (DNN), were employed to analyze the data. Key features influencing BR prediction included prostate capsule infiltration, postoperative Gleason score, surgical margins, and tumor volume. The dataset was divided into 70% for training and 30% for testing, with an external validation cohort of 96 patients. The performance of the models was evaluated using the area under the receiver operating characteristic curve (AUC). The XGBoost model performed the best, with an AUC of 0.91, classifying 84% of the cases accurately, followed closely by RF (AUC 0.90) and DNN (AUC 0.89). The best-performing model, XGBoost, was externally validated and showed an AUC of 0.90, with an accuracy of 81%. This was superior to the traditional CAPRA-S model, which had an accuracy of 77% and AUC of 0.77. These results suggest that AI models, particularly XGBoost, provide a more accurate prediction of BR compared to traditional statistical models. The ability of AI to analyze complex data patterns could enable more precise and personalized treatment planning, ultimately improving patient outcomes by identifying those at higher risk of recurrence early. AI-driven models offer a significant advancement in the management of prostate cancer, supporting clinicians in making informed, evidence-based decisions.