Machine learning for UAS insertion in ureteroscopy: role of distal ureteral angle

This study aims to identify key factors influencing the success rate of ureteral access sheath (UAS) insertion during flexible ureteroscopy (FURS) and to develop a predictive model using machine learning (ML) techniques to optimize surgical procedures and improve insertion success rates. A retrospective analysis was conducted on 307 patients with upper ureteral or renal stones treated at the Department of Urology, Xiangya School of Medicine, Central South University, between January 2022 and August 2023. Based on UAS insertion outcome, patients were classified into success and failure groups. Logistic regression identified factors associated with successful UAS insertion. Ten different machine learning models were trained to predict UAS insertion success, with performance evaluated using metrics such as AUC, accuracy, recall, and F1 score. SHapley Additive Explanations (SHAP) were applied to interpret the models and identify the most influential factors. Of 307 patients, 217 (70.7%) successfully underwent UAS insertion. Logistic regression identified the angle between the distal ureter and the body’s longitudinal axis (Angle) as a significant independent factor ( P < 0.001, OR: 1.60, 2.82), with larger angles associated with higher success rates. Among the ML models, CatBoost outperformed others, achieving an AUC of 1.0 in the training set and 0.9892 in the test set. Its accuracy, recall, and other metrics were superior. SHAP analysis confirmed that the Angle was the most influential factor in predicting UAS success. The angle between the distal ureter and the body’s longitudinal axis is a key factor in UAS insertion success. The CatBoost model demonstrated excellent predictive performance and may offer valuable insights for developing personalized surgical strategies in the future.