摘要
为了更细节地提取手部X图像特征,提出了一种基于改进残差网络的骨龄评估方法。基于TW3方法,结合轻量高效注意模块改进残差网络,提高细小颗粒特征的提取准确率。实验结果表明,该方法在西安某三甲医院提供的数据集上,男、女性的平均绝对误差(MAE)分别是0.4228岁和0.4341岁,在1岁误差范围内,男、女性的准确率分别达到93.82%和93.16%,明显地提高了骨龄评估的准确率。
In order to extract features from hand X images in more detail,this paper proposes a method of bone age assessment based on improved residual network.Based on the TW3 method,combined with lightweight and efficient attention modules,it improves the extraction accuracy of minuscule particle features.Experimental results demonstrate that the method has the Mean Absolute Error(MAE)of 0.4228 years and 0.4341 years for males and females on the dataset provided by a tertiary hospital in Xi'an.Within the one-year error range,the accuracy of males and females reaches 93.82%and 93.16%,respectively,significantly improving the accuracy of bone age assessment.
作者
马瑞齐
胡晓丹
席秀蕾
MA Ruiqi;HU Xiaodan;XI Xiulei(Zhengzhou University of Science and Technology,Zhengzhou 450064,China)
出处
《现代信息科技》
2024年第15期134-137,141,共5页
Modern Information Technology
关键词
骨龄评估
改进残差网络
轻量高效注意模块
TW3
bone age assessment
improved residual network
lightweight and efficient attention module
TW3