Fracture Detection Health Network (FDHN): A solution to generate bone fracture insight

This research addresses the critical issue of bone fracture detection in medical imaging using FDHN, emphasizing the need for efficient and reliable techniques to enhance patient care. Leveraging the YOLOv8 real-time object detection model, the study focuses on automating and optimizing the bone fracture detection process in X-ray images. The architecture of YOLOv8, with its balanced performance characteristics, is explored, and a custom dataset of 200 annotated X-ray images is employed for model training and testing. The study demonstrates the impact of training duration on model accuracy, achieving a mean Average Precision (mAP) of 0.886 at the 50 IoU (Intersection over Union) threshold after 100 epochs. The model's ability to detect objects under stricter criteria (50-95 IoU) also significantly improved, reaching a mAP of 0.636. These results underscore the value of extended training in enhancing the model's accuracy, making it a reliable tool for bone fracture detection in medical imagery. The research contributes to optimizing clinical workflows, reducing diagnostic errors, and ultimately improving patient outcomes in the healthcare sector.