BUILDING AN AUTOMATED MODEL FOR SPERM RECOGNITION AND QUALITY ASSESSMENT BASED ON HALO IMAGES

Abstract

The fundamental challenge in the diagnosis and treatment of infertility is the need to perform a quick and accurate analysis, assessment, and classification of sperm quality. There are various methods for assessing the level of sperm DNA fragmentation. The most common method currently is the evaluation of sperm chromatin dispersion (SCD) developed by Fernandez and colleagues. This method is simple, cost-effective, and can be easily carried out using a conventional microscope, making it readily applicable in clinical settings. However, the classification of sperm in halo images remains challenging due to the diversity in sperm shapes and sizes. Therefore, there is a need for new methods to standardize, automate, and expedite the sperm classification process. This paper presents the use of the RetinaNet model for detecting and classifying sperm in halo images. Experimental results show that the RetinaNet model achieves good accuracy in detecting and classifying sperm in halo images. The paper also discusses the challenges and future directions in this field.