Machine Learning-Based Prediction of Urolithiasis Recurrence Using Patient's Clinical Data, Demography, and CT...

Area of Science:

• Nephrology
• Medical Informatics
• Data Science

Background:

• Urolithiasis, or urinary tract stone formation, requires timely diagnosis for effective treatment and complication prevention.
• Predicting recurrence is crucial for managing patients with a history of kidney stones.

Purpose of the Study:

• To develop and evaluate machine learning (ML) models for predicting urolithiasis recurrence.
• To identify the most effective ML classifier for urolithiasis recurrence prediction using clinical, demographic, and CT data.

Main Methods:

• Utilized clinical data, demographics, and CT findings from 4246 patients over three years.
• Developed and evaluated six ML classifiers, including random forest, using train/test split and k-fold cross-validation.
• Assessed model performance using ROC curve analysis, calibration analysis, and decision curve analysis.

Main Results:

• Random forest demonstrated the best performance, achieving an Area Under the ROC Curve (AUC) of 0.64 with train/test split.
• K-fold cross-validation showed random forest with an AUC of 0.63, sensitivity of 0.90, and positive predictive value of 0.83.
• The random forest model exhibited good calibration with a Brier score of 0.18.

Conclusions:

• Machine learning offers a practical approach to predicting urolithiasis recurrence with clinically acceptable accuracy.
• The study highlights the potential of ML models, particularly random forest, to aid in clinical decision-making for urolithiasis management.
• Evaluated ML models provide a more accurate prediction compared to traditional scoring systems for urolithiasis recurrence.