A Comparative Analysis of YOLOv8, YOLOv9, YOLOv10 and YOLOv11 for Bone Fracture Detection

Abstract

Early diagnosis of bone fractures is critical for treatment success, and collaboration between radiologists and surgeons plays a crucial role in this process. The use of advanced technologies accelerates the diagnostic process, resulting in improved patient health outcomes. This study aims to accurately and effectively detect bone fractures in the FracAtlas dataset using the latest variants of the YOLO architecture. The integration of deep learning architectures with data augmentation techniques improves the accuracy and reliability of fracture detection. In the study results, the best performance was achieved by the YOLOv10x model combined with automatic data augmentation techniques, yielding a precision of 0.882, recall of 0.758 and mAP@50 of 0.8016. The findings of this study underscore the potential of AI-powered fracture detection systems to improve clinical decisions and expedite fracture diagnosis delays. © 2025 IEEE.