Automated Detection of Urine Sediments using YOLOv8

Urine sediment analysis, a critical clinical technique that is frequently conducted manually with a microscope, has seen minimal progress in terms of workflow automation. This study describes an automated approach for recognizing and classifying urine sediment particles (red blood cells, white blood cells, epithelial cells, crystals, and casts) using the YOLOv8 object detection algorithm. Advanced feature extraction approaches (LBP and HOG) and considerable preprocessing (denoising, color correction) were used to improve detection. At an IoU threshold of 0.5, the system attained an average precision (mAP) of 87.4 percent. Future development will concentrate on integrating the system with mobile platforms and laboratory information systems to improve clinical processes by minimizing human error and diagnosis time.