摘要
在预测成年身高和诊断生长障碍时,骨龄是重要判断指标之一。参照骨的准确提取是保证骨龄评估准确性的关键。然而,现有的方法会导致部分参照骨的漏判和误判,从而降低14个参照骨的平均提取精度。基于优化的YOLO算法,结合可变卷积、坐标注意力、特征层扩展以及EIOU损失,提出了一种手腕部X光片参照骨提取方法,IM-YOLO。该方法可以更好地提取不规则参照骨的特征信息,提高小关节间隙区域的检测能力,将参与CHN法骨龄评估的14个参照骨准确地提取出来。实验结果表明,IM-YOLO算法检测速度快,精度高,所有参照骨的检测精度都达到99%以上,FPS指数为16。
Bone age is one of the important indicators for predicting adult height and diagnosing growth diseases.The accurate extraction of reference bone is the key to ensure the accuracy of bone age assessment.However,existing methods lead to missed and misjudged reference bones.Thus,the average extraction accuracy of the 14 reference bones is reduced.A wrist bone reference bones extraction method named IM-YOLO is proposed.This method is based on optimized YOLO algorithm and combined with deformable convolution,coordinate attention,feature-level expansion and EIOU loss.IM-YOLO can better extract the feature information of irregular reference bones,improve the detection ability of facet joint space area,and accurately extract 14 reference bones.The experimental results show that the IM-YOLO algorithm has fast detection speed and high accuracy.The detection accuracy of all reference bones is over 99.8%,and the FPS is 16.
作者
周卫元
武坤秀
张拓
毛严
武佳男
毛科技
何文秀
ZHOU Weiyuan;WU Kunxiu;ZHANG Tuo;MAO Yan;WU Jianan;MAO Keji;HE Wenxiu(College of Xiaoshan,Zhejiang Open University,Hangzhou Zhejiang 312000,China;College of Computer Science and Technology,Zhejiang University of Technology,Hangzhou Zhejiang 310023,China;Zhijiang College,Zhejiang University of Technology,Shaoxing Zhejiang 312030,China)
出处
《传感技术学报》
CAS
CSCD
北大核心
2023年第9期1426-1435,共10页
Chinese Journal of Sensors and Actuators
基金
国家自然科学基金项目(62072410)
浙江省基础公益研究计划项目(LTGG23F020002,LGG22F020014,LGF21F020015)。
