[1]王荣翔,赵静文.基于改进YOLOv8的轻量级腕部X光骨折检测算法[J].中国医学物理学杂志,2025,42(6):740-750.[doi:DOI:10.3969/j.issn.1005-202X.2025.06.006]
　WANG Rongxiang,ZHAO Jingwen.A lightweight wrist fracture detection algorithm based on an enhanced YOLOv8 model forX-ray imaging[J].Chinese Journal of Medical Physics,2025,42(6):740-750.[doi:DOI:10.3969/j.issn.1005-202X.2025.06.006]
点击复制

基于改进YOLOv8的轻量级腕部X光骨折检测算法()

《中国医学物理学杂志》[ISSN:1005-202X/CN:44-1351/R]

卷:
42
期数:
2025年第6期
页码:
740-750
栏目:
医学影像物理
出版日期:
2025-06-20

文章信息/Info

Title:
A lightweight wrist fracture detection algorithm based on an enhanced YOLOv8 model forX-ray imaging
文章编号:
1005-202X（2025）06-0740-11
作者:
王荣翔; 赵静文
上海工程技术大学电子电气工程学院，上海 201620
Author(s):
WANG Rongxiang;  ZHAO Jingwen
School of Electronic and Electrical Engineering, Shanghai University of Engineering Science, Shanghai 201620, China
关键词:
腕部骨折检测; YOLOv8; 轻量化模型; 深度学习; 医疗影像分析
Keywords:
wrist fracture detection;  YOLOv8;  lightweight model;  deep learning;  medical image analysis
分类号:
R318；TP391
DOI:
DOI:10.3969/j.issn.1005-202X.2025.06.006
文献标志码:
A
摘要:
为提升腕部X光影像中骨折检测的精度与效率，提出一种基于改进YOLOv8的轻量级目标检测算法YOLOv8-DLE。该算法在原有YOLOv8框架的基础上，结合多尺度特征提取方法（DWR）、大型分离卷积注意力模块（LSKA）以及高效卷积优化模块（EfficientRepBiPAN）。在有效减少计算量的同时，显著提升模型对小目标的检测能力和复杂背景处理能力。实验结果表明，YOLOv8-DLE 在 GRAZPEDWRI-DX 数据集上的 mAP@50 较原始 YOLOv8 提升 3.7%，mAP@50:95 提升 1.7%，YOLOv8-DLE 在保持较高检测精度的基础上，参数量从 11.1 M 减少至 10.9 M，降低 0.2 M，GFLOPs从28.5减少至26.0，降低2.5。此外，YOLOv8-DLE在嵌入式设备和远程医疗系统中具有良好的适应性，特别适用于资源受限的环境，能够为医生提供实时辅助诊断支持，帮助提升诊断的准确性与效率。该模型的轻量化设计和高效性为未来在医学影像分析领域的广泛应用提供新的可能性。
Abstract:
A lightweight object detection algorithm based on an enhanced YOLOv8 framework (YOLOv8-DLE) is proposedto improve the accuracy and efficiency of wrist fracture detection in X-ray images. On the basis of original YOLOv8framework, the model integrates a dilation-wise residual module, a large separable kernel attention module and anEfficientRepBiPAN for improving the model’s ability to detect small targets and manage complex backgrounds whilereducing computational cost. Experimental results on the GRAZPEDWRI-DX dataset demonstrate that YOLOv8-DLEoutperforms the original YOLOv8, achieving a 3.7% increase in mAP@50 and a 1.7% increase in mAP@50: 95, withreductions in parameters from 11.1 M to 10.9 M and GFLOPs from 28.5 to 26.0. The model’s compactness and efficiencymake it well-suited for embedded devices and remote healthcare systems, particularly in resource-limited environments.YOLOv8-DLE can provide real-time auxiliary diagnostic support for doctors and improve the accuracy and efficiency ofdiagnosis, showing strong potential for real-time clinical deployment and broad applicability in medical image analysis.

相似文献/References:

[1]乐振,孙振,鞠瑞文,等.基于YOLOv8s的改进结核病病原体检测算法[J].中国医学物理学杂志,2024,41(7):910.[doi:DOI:10.3969/j.issn.1005-202X.2024.07.019]
　YUE Zhen,SUN Zhen,JU Ruiwen,et al.A novel tuberculosis pathogens detection algorithm based on YOLOv8s[J].Chinese Journal of Medical Physics,2024,41(6):910.[doi:DOI:10.3969/j.issn.1005-202X.2024.07.019]
[2]彭志博,陈勇,崔艳荣.基于YOLOv8m的改进腕部X光片骨折检测算法[J].中国医学物理学杂志,2025,42(4):542.[doi:10.3969/j.issn.1005-202X.2025.04.017]
　PENG Zhibo,CHEN Yong,CUI Yanrong.Fracture detection in wrist X-ray image using an improved algorithm based on YOLOv8m[J].Chinese Journal of Medical Physics,2025,42(6):542.[doi:10.3969/j.issn.1005-202X.2025.04.017]

备注/Memo

备注/Memo:
【收稿日期】2025-01-13【基金项目】国家自然科学基金（62303303）【作者简介】王荣翔，硕士研究生，研究方向：医学图像处理，E-mail:2532047741@qq.com【通信作者】赵静文，讲师，研究方向：医学图像处理，E-mail: jingwen_echo@outlook.com
更新日期/Last Update: 2025-06-30