Efficient YOLOv11-Based Approach with Dual-Path Feature Learning for Automated Detection of Limb Fractures

Abstract

The proper detection of limb fractures by use of radiographic images is crucial to medical institutions that have low resources. The study proposes a new deep learning platform named YOLOv11 that enhances the speed and accuracy of bone fracture tailoring in medical images. The YOLOv11 system has identified the weaknesses of the earlier versions of the YOLO version by having a dual-path solution to feature extraction and the mechanism of improved attention and gradient consistency refinement. Two publicly available radiograph databases were utilised by the research to construct a hybrid dataset comprising 4,739 labelled images and 1,030 X-rays of Gujranwala Medical College Hospital to be used diversely and practically. The evaluation results showed YOLOv11 achieved 0.89 precision and 0.81 mAP@0.5 and 0.55 mAP@0.5–0.95 while outperforming YOLOv8 and YOLOv10 and Faster R-CNN in both performance and speed performance. The model showed excellent performance on local clinical data and it processed images at 62 FPS in real-time. YOLOv11 serves as an essential tool for AI-based radiology support in orthopedic trauma treatment because it provides high diagnostic performance with minimal system requirements.

Ähnliche Arbeiten