Using of YOLO algorithm for detection of particles in urine sediment

Urine produced by the kidney and passing through the urinary tract consists of insoluble wastes. These wastes contain important information for urinary tract and kidney diseases. Diagnosis and diagnosis of the disease is made by urine analysis. Complete Urine Analysis is performed by performing physical, chemical and microscopic examination. Microscopic examination is very important for definitive results. In microscopic examination, a microscopic image of the urine sediment is taken and the particle types in the image are detected by their visual forms and their numbers are determined to diagnose and diagnose the disease. The traditional method of manual examination is used in microscopic examination of urine. However, this method requires expert knowledge and is a time-consuming process. At the same time, detection of particles in complex urine images can sometimes be difficult and may produce non-standardized and incorrect results. Systems and software have been developed to automate measurements and tests to overcome these and similar difficulties. The aim of the study is to automatically detect particles in urine sediment with accurate, consistent and standardized results. For this purpose, a YOLO11-based method has been proposed. Eight urine particle types were detected, including red blood cells, white blood cells, epithelium, bacteria, yeast, crystals, cylinders and others. The proposed method used YOLO11n, YOLO11s, YOLO11m, YOLO11l and YOLO11xl variants of the YOLO11 model. The results were compared with the obtained performance metrics. Due to the size of the dataset, the best results were obtained in the YOLO11s model and a mAP score of 79.8% was obtained. The proposed method and the obtained results show that it can be used both as an auxiliary visual support and as an educational material for physicians in the microscopic examination of urine sediment.