Screening for Clinically Significant Nephrolithiasis Based on Simple Health Checkup Clinical and Urine Parameters in...

Area of Science:

• Nephrology
• Artificial Intelligence
• Medical Diagnostics

Background:

• Nephrolithiasis affects 15% of the population, often undetected in asymptomatic individuals.
• Current imaging diagnostics are costly, operator-dependent, or involve radiation.
• A practical, low-cost screening strategy for kidney stones is needed.

Purpose of the Study:

• To develop a low-cost, rapid screening model for clinically significant nephrolithiasis.
• Utilize machine learning (ML) and simple clinical parameters for early kidney stone detection.

Main Methods:

• Retrospective study of 6528 adults across 3 hospitals in Taiwan (2012-2021).
• Trained ML models using 10 routine variables including demographics, medical history, and urine analysis.
• Evaluated models using ROC and PR curves; Shapley value analysis for interpretability.

Main Results:

• The best ML model achieved an AUC of 0.968 and AUPRC of 0.936.
• Demonstrated high sensitivity (87.3%) and specificity (94.7%) for kidney stone detection.
• Key predictors identified: urine red blood cell count, estimated glomerular filtration rate, and urine specific gravity.

Conclusions:

• Developed an efficient, noninvasive ML model for large-scale kidney stone screening using routine health data.
• Model can integrate into health checkups and telemedicine for proactive management.
• Further validation in diverse populations is recommended for generalizability.