Comparative Analysis of Object Detection Frameworks for Fracture Detection in X-Ray Image

Fracture detection plays a critical role in clinical examinations, especially in emergency surgery. Traditional fracture diagnosis relies on the experience of radiologists, which carries the risk of misdiagnosis. With the advancement of deep learning technologies, object detection methods have been widely applied to automated fracture detection, providing efficient and accurate solutions. This study aims to evaluate the performance of various object detection frameworks in the task of fracture detection by comparing one-stage and two-stage detectors, anchor-based and anchor-free methods. Thus, we selected nine representative object detection models for comparison, covering a variety of deep-learning architectures. Experimental results show that YOLOv10 not only achieves the highest accuracy but also demonstrates significant advantages in inference speed. Furthermore, Transformer-based models exhibit better precision in fracture detection, particularly showing potential in recognizing complex image features.