Artificial intelligence model for predicting early biochemical recurrence of prostate cancer after robotic-assisted radical prostatectomy

Abstract Prostate cancer remains a significant public health concern, with a substantial proportion of patients experiencing biochemical recurrence (BCR) after radical prostatectomy (RP). Traditional risk models, such as CAPRA-S, have demonstrated moderate predictive performance, highlighting the ne...

Full description

Saved in:
Bibliographic Details
Main Authors: Miguel Angel Bergero, Pablo Martínez, Patricio Modina, Ricardo Hosman, Wenceslao Villamil, Romina Gudiño, Carlos David, Lucas Costa
Format: Article
Language:English
Published: Nature Portfolio 2025-08-01
Series:Scientific Reports
Subjects:
Online Access:https://doi.org/10.1038/s41598-025-16362-1
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Abstract Prostate cancer remains a significant public health concern, with a substantial proportion of patients experiencing biochemical recurrence (BCR) after radical prostatectomy (RP). Traditional risk models, such as CAPRA-S, have demonstrated moderate predictive performance, highlighting the need for more accurate tools. This study aimed to develop a machine learning (ML) model to predict BCR in patients undergoing robot-assisted laparoscopic RP (RALP). A retrospective cohort of 1024 (476 BCR+ and 548 BCR−) patients was analyzed, using a balanced dataset of 25 clinical and pathological variables. Five ML classifiers were evaluated, with XGBoost emerging as the best-performing model, achieving 84% accuracy and an AUC of 0.91. Model validation on an independent dataset of 96 patients confirmed its robustness, with an AUC of 0.89. Decision and calibration curves demonstrated the superior clinical applicability of XGBoost compared to CAPRA-S, indicating improved risk stratification and potential to optimize treatment decisions. The study underscores the value of ML in refining prognosis prediction and guiding therapeutic strategies in prostate cancer. While further validation in diverse clinical settings is necessary, these findings support the integration of ML-based models into clinical decision-making to enhance personalized patient management.
ISSN:2045-2322