Failed placement of a Ureteral Access Sheath (UAS) during flexible ureteroscopic lithotripsy (FURL) creates significant operative challenges. The objective of this study was to develop a machine learning (ML) model that integrates diverse indicators to predict the likelihood of UAS placement failure. We conducted an analysis of clinical records from 1,153 patients diagnosed with upper urinary tract calculi, followed by the enrollment of 265 consecutive patients for independent prospective validation. The predictive model integrated variables including patient medical history, parameters from CT imaging, anesthesia modality, and markers of systemic inflammation. Nine distinct ML algorithms were formulated. The predictive performance of the model was subsequently compared with that of senior urologists in a prospective setting. The final dataset comprised retrospective (n = 651) and prospective (n = 265) cohorts, both exhibiting similar failure rates of approximately 30%. Among the tested algorithms, the RF model yielded superior performance, attaining an area under the curve (AUC) of 0.845 in the testing set. During prospective validation, the simplified RF model achieved an AUC of 0.872, significantly surpassing the predictive accuracy of senior urologists. Through SHAP analysis, the mid-ureteral diameter, distal ureteral diameter, neutrophil count, renal parenchyma width, and the long diameter of the calculus were identified as the five most critical predictors. Furthermore, the model exhibited robust stability across various subgroups. We successfully developed and prospectively validated a robust ML model capable of predicting UAS placement failure. By assimilating clinical, imaging, and inflammatory data, this model demonstrated superior performance compared to clinical experts. It provides a dependable instrument for personalized preoperative risk stratification and surgical strategy planning, accessible via a newly deployed web application.
Li et al. (2026) studied this question.