Integrated Urine Crystal Detection and Classification Using Sensor Data and YOLOv8 Imaging

Kidney stones are a prevalent and painful urological condition affecting up to 10% of the global population, particularly men aged 30 to 60. Early detection is essential to prevent complications such as nephrolithiasis and chronic kidney disease. This study presents an integrated system that combines chemical and imaging-based urine analysis to detect and classify urine crystals while measuring pH and turbidity. The system utilizes a pH sensor (PH-4502C) and a turbidity sensor (SENO189), whose outputs are interpreted through logistic regression. Imaging is conducted using a Raspberry Pi High-Quality Camera mounted on a microscope, and crystal classification is performed using the YOLOv8 object detection model trained on labeled datasets of calcium oxalate and triple phosphate crystals. All modules are operated through a Raspberry Pi 4 Model B. Validation with 30 clinical urine samples demonstrated high concordance with standard laboratory results. The system achieved an R² value of 0.983 for pH detection, 96.67% accuracy in turbidity classification, and an R² of 0.976 with a mean absolute error of 0.733 for crystal counting. Overall risk assessment achieved 96.67% accuracy, 100% precision, 88.89% recall, and an F1 score of 94.12%. These results confirm the system’s accuracy, reliability, and suitability for practical use. It offers a low-cost, noninvasive, and real-time solution for early kidney stone detection, with strong potential for application in point-of-care diagnostic settings.