An artificial intelligence-based clinical decision support system for large kidney stone treatment

Area of Science:

• Urology
• Medical Informatics
• Machine Learning

Background:

• Percutaneous nephrolithotomy (PCNL) is a common procedure for kidney stone treatment.
• Predicting postoperative outcomes and complications is crucial for patient management and surgical planning.

Purpose of the Study:

• To develop and evaluate a decision support system (DSS) for predicting postoperative outcomes of kidney stone treatment, specifically PCNL.
• To enhance preoperative counseling and surgical decision-making for urologists.

Main Methods:

• Collected pre/postoperative data from 254 patients, including patient history, kidney stone parameters, and laboratory data.
• Developed a prediction model using machine learning, incorporating dimensionality reduction (sequential forward selection and Fisher's discriminant analysis) and a multiple classifier scheme.
• Validated the DSS using a leave-one-patient-out cross-validation approach.

Main Results:

• The DSS achieved high accuracy (94.8%) in predicting overall treatment outcomes.
• The system correctly estimated the need for stent placement in 85.2% of cases.
• The DSS accurately predicted the requirement for blood transfusion in 95.0% of cases.

Conclusions:

• The developed DSS is a promising tool for assisting urologists in predicting surgical outcomes for kidney stone removal.
• The system can improve patient counseling and aid in the selection of appropriate surgical treatments.
• The integration of machine learning in DSS can significantly enhance the precision of predicting postoperative results in urological procedures.