Comparative Performance Analysis of YOLOv8, YOLOv9, and YOLOv10 Models in Detecting Human Auditory Hair Cells

This study is centered on enhancing the detection of auditory hair cells in the human inner ear employing deep learning techniques. It compares the analysis of inner and outer auditory hair cells from high-resolution images acquired using advanced imaging modalities, including confocal and fluorescence microscopy among three YOLO family models—YOLOv8, YOLOv9, and YOLOv10. The methodology incorporates three sections—data collection and preparation; annotation using CVAT for manual labeling ear cell images; and model training and testing utilizing YOLO models. Experimental analysis established that YOLOv10 provided an overall increased detection accuracy of 81.4%, followed by YOLOv8 and YOLOv9 with sequential accuracy of 64.4%. As a result, YOLOv10 was more reliable than YOLOv8 and YOLOv9 for detecting auditory hair cells in the inner ear. Therefore, this research confirmed the YOLOv10 model as an increased automated annotation. And place it into an existing detection system increasing diagnostics in auditory science. Future analysis will focus on additional optimization of the YOLOv10 model and application in clinical and research studies that will be beneficial for diagnosis and treatment of hearing disorders.