Vocal Folds Image Segmentation Based on YOLO Network

The focus of this article is on utilizing YOLOv8 segmentation models for the detection of vocal fold openness in laryngoscopic videos, eliminating the need for extra image enhancement. The evaluation and comparison of different models are carried out based on accuracy metrics such as box mean average precision and mask mean average precision. The outcomes indicate the potential applicability of YOLOv8 segmentation models in objectively quantifying vocal fold openness, offering a potential avenue for integration into clinical practice.