Urine Calcium Oxalate Crystallization Recognition Method Based on Deep Learning

As a global disease, urolithiasis has a high incidence and recurrence rate, which can seriously threaten patients' life and physical and mental health. Calcium oxalate crystals are the major constituents of stones. The analysis of calcium oxalate crystals in urine has important clinical significance. The traditional identification methods include regional growth method, edge detection method, activity model method and mathematical morphology method. These methods have higher requirements for the identified urine sediment microscopy images, and the recognition accuracy rate is not high. In this paper, a novel identification method based on full convolution neural network is proposed, which consists of resnet50, FPN and a classification subnet. The recognition rate is significantly improved to 74% in our experiments.