Flexible ureteroscopy in renal anomalies: an explainable AI model for surgical outcome prediction from EAU endourology

Aim and objective Flexible ureteroscopy (fURS) is a well-established modality for managing urolithiasis in patients with congenital renal anomalies such as horseshoe kidneys (HK), malrotated kidneys (MK), and pelvic ectopic kidneys (PEK). Still, these anatomical variants present unique challenges that complicate stone clearance and procedural planning. We aim to apply machine learning (ML) and explainable artificial intelligence (XAI) techniques to identify predictors of stone-free status (SFS) following fURS in patients with anomalous kidneys. Methods We retrospectively analysed adult patients with HK, MK, or EK who underwent fURS and laser lithotripsy for renal stones at a tertiary referral center. A ML model incorporating clinical and intraoperative variables was developed to predict SFS. SHAP (SHapley Additive exPlanations) values and decision tree analysis were used to interpret feature impor- tance and model behaviour. Results A total of 569 cases were analysed between 2017 and 2021, with a female: male ratio of 3:1. Regarding anatomical anomalies, 50.62% had HSK, 22.67% had PEK and 26.71% had MK. Most of the patients presented with multiple (59.58%), small (76.80%) and soft stones (56.94%). MK showed the highest SFS rates, suggesting this is the most favourable anomaly for fURS. The presence of residual fragments at the end of the procedure was the strongest negative predictor of SFS, fol- lowed by longer operative time and older patient age. PEK exhibited the greatest heterogeneity in outcomes. SHAP analysis provided individualized and global insights into feature contributions. Conclusion Explainable AI offers a transparent and clinically meaningful approach to predicting SFS in patients with renal anomalies undergoing fURS. These insights can guide preoperative risk stratification and inform surgical strategy in a domain where standardised evidence is lacking.