YOLO-Based Detection for Sperm Motility Analysis

“Motility” refers to the ability of an organism or liquid to move. In the context of sperm, it pertains to the movement and swimming ability of sperm cells, which is a crucial factor in sperm health assessed during sperm analysis. Poor sperm motility indicates that sperm are not swimming properly, which can contribute to male infertility. Despite being available for 25 years, computer-aided sperm analysis (CASA) systems still face significant technical challenges as automated semen analyzers. The primary challenge is accurately determining sperm concentration in semen, which affects the ability to determine the percentage of motile spermatozoa. These issues need to be resolved to improve the reliability and effectiveness of CASA systems. This study evaluated the performance of the latest object detection architectures, YOLOv8 and YOLO-NAS, for sperm detection. Experimental results on the VISEM dataset showed that the YOLOv8 model achieved mAP50 and recall rates of 99.27% and 98.33%, respectively, outperforming the YOLO-NAS model, which achieved 92.03% and 98.15%, respectively.