Human Sperm Morphology Classification Using YOLOv5 Deep Learning Algorithm

Over the past two decades, one of the most common reproductive problems has been infertility. Infertility is a disease of the reproductive system characterized by the inability to conceive after 12 months or more of regular unprotected sexual intercourse. According to the World Health Organization (WHO), the reproductive rate has declined drastically, with male infertility now accounting for 36 percent of cases, often due to abnormalities in sperm production. Currently, infertility screening is still done manually, evaluating sperm samples using a microscope, which often produces inconsistent results. However, advances in computer technology have resulted in significant research aimed at improving the analysis of male sperm infertility. This study utilized deep learning technology to identify sperm using the YOLOv5 method. This study involved several stages with data collection using 1330 photos with 2 classes, namely sperm and non-sperm in video format. The second stage involved preprocessing the dataset, which included data extraction, cropping, resizing, and labeling the data for training. The final stage involved testing the trained model to detect and classify sperm based on morphology. The experimental results show that the proposed method is effective in accurately classifying sperm images and analyzing motion videos from recorded video data with a mAP result of 73.1 percent.Therefore, in the context of this study, the YOLOv5 model is considered efficient in detecting and classifying objects, both sperm and non-sperm.