Design and Implementation of YOLOv11-NWMD: A Lightweight Object Detection Algorithm for X-ray Medical Imaging

Zhipan Wu, Zirui Li, Guoming Lai, Junjie Ji, Huaying Du, Mian Zeng, Jinxiong Chen

Keywords: Fracture Detection; YOLOv11 Improvement; Dynamic Feature Fusion; Multi-Frequency Multi-Scale Attention; Medical Imaging Object Detection

TL;DR: Design and Implementation of improved YOLOv11-NWMD

Abstract:

This study aims to improve the accuracy and efficiency of fracture detection in X-ray images. A lightweight improved model, YOLOv11-WMD,
based on YOLOv11-n is proposed. To address challenges commonly encountered in medical imaging, such as noise interference, insufficient multiscale feature fusion, limited receptive field, and poor small target detection performance, the model integrates a Multi-Frequency Multi-Scale Attention (MFMSA) module and a Dynamic Feature Fusion (DFF) module. Experimental results show that while
maintaining a training time comparable to the original YOLOv11-n, YOLOv11-NWMD achieves a significant improvement in precision (Precision = 0.953, 0.7% higher than the baseline model YOLOv11-n) and a breakthrough increase in recall (Recall = 0.904, 5.4% higher than YOLOv11-n), significant enhances training efficiency. This verifies the effectiveness of YOLOv11-NWMD in achieving high model performance.

Submission Number: 11