Ju Rui-Yang ; Chien Chun-Tse ; Lin Chia-Min ; Chiang Jen-Shiun

Abstract

Children often suffer wrist injuries in daily life, while fracture injuringradiologists usually need to analyze and interpret X-ray images before surgicaltreatment by surgeons. The development of deep learning has enabled neuralnetwork models to work as computer-assisted diagnosis (CAD) tools to helpdoctors and experts in diagnosis. Since the YOLOv8 models have obtained thesatisfactory success in object detection tasks, it has been applied to fracturedetection. The Global Context (GC) block effectively models the global contextin a lightweight way, and incorporating it into YOLOv8 can greatly improve themodel performance. This paper proposes the YOLOv8+GC model for fracturedetection, which is an improved version of the YOLOv8 model with the GC block.Experimental results demonstrate that compared to the original YOLOv8 model,the proposed YOLOv8-GC model increases the mean average precision calculated atintersection over union threshold of 0.5 (mAP 50) from 63.58% to 66.32% on theGRAZPEDWRI-DX dataset, achieving the state-of-the-art (SOTA) level. Theimplementation code for this work is available on GitHub athttps://github.com/RuiyangJu/YOLOv8_Global_Context_Fracture_Detection.