Deep Learning-based Automated Assessment of Sperm Motility

Assessing sperm motility is one of the essential factors in semen analysis and infertility diagnosis. Traditional methods take time and require specialized skills for accurate assessment. However, with the development of artificial intelligence especially deep learning techniques, the ability to automatically assess sperm motility has opened up. In this study, we present a method to evaluate the motility parameters of sperms based on deep learning techniques with the VISEM data set. In addition to using Mean Absolute Error (MAE) and Mean Squared Error (MSE), the loss function is also extended to proper to the proposed investigation. The method achieves the best results with an MAE of 9.322 and 9.620 of extended MAE. The results demonstrate the feasibility of predicting sperm motility based on deep learning, and that it may become a rapid survey and assessment tool for sperm motility evaluation.