A novel predictive method for URS and laser lithotripsy using machine learning and explainable AI: results from the...

Area of Science:

• Urology
• Medical Informatics
• Artificial Intelligence

Background:

• Ureteroscopy (URS) is a common procedure for treating urolithiasis (kidney stones).
• Predicting outcomes of URS can significantly improve patient care and resource allocation.
• Current prediction methods may lack the precision needed for personalized treatment planning.

Purpose of the Study:

• To develop and validate machine learning (ML) algorithms for predicting ureteroscopy (URS) outcomes.
• To identify key factors influencing stone-free status and postoperative complications.
• To enhance clinical decision-making and personalized care in urolithiasis management.

Main Methods:

• Utilized the FLEXOR database, comprising 6669 patients undergoing URS for urolithiasis (2015-2023).
• Trained 15 ML algorithms to analyze preoperative and postoperative data, correlating them with outcomes.
• Employed Explainable AI to identify critical features influencing prediction accuracy.

Main Results:

• Extra Tree Classifier achieved 81% accuracy in predicting stone-free status (SFS).
• Identified significant correlations between various clinical factors (e.g., stone characteristics, medications, scope type) and outcomes like postoperative bleeding, injury, and fever.
• ML models effectively predicted SFS, influenced by factors such as age, stone diameter, and specific surgical tools.

Conclusions:

• Machine learning offers a valuable tool for accurately predicting URS outcomes using existing patient data.
• The developed ML models demonstrate strong performance in predicting outcomes and risks.
• This research establishes a foundation for creating accessible predictive tools to guide clinical practice in urolithiasis treatment.