摘要
骨龄是儿童青少年生长发育的重要指标,目前自动化骨龄判读算法研究中,针对骨化中心的准确分割提取是重要一环。然而之前的研究多采用单一算法、过度追求算法效率,导致稳定性、准确度不高。基于《中国人手腕骨发育标准-中华05》中RUS-CHN方法,针对13处骨化中心的提取问题,改进了骨化中心提取的准确度和稳定性。所用算法包括Canny边缘识别、基于局部图像灰度特征的边缘筛选,边缘连接、骨中轴线多项式拟合、基于Gabor纹理分析的骨化中心精确定位算法。实验表明:对随机抽样的980张待分割样本,应用综合算删除法,对RUS-CHN判读方法所关注的13个骨化中心区域进行分割,13个骨化中心均正确识别并分割的成功率达到了99.2%。
Bone age is an important index for the growth and development of children and adolescents. In the study of automated bone age reorganization algorithms,accurate segmentation and extraction of ossification centers is an important link. However,previous studies used single algorithm and excessively pursued algorithm efficiency,resulting in low stability and accuracy. Based on the RUS-CHN method in The Standards of Bone Age in Hand and Wrist for Chinese： China 05,this study focused on the extraction of ossification centers,and improved the accuracy and stability. The algorithm in this study included Canny edge recognition,edge filter based on grayscale,edge linking,polynomial fit,center searching and center recognition based on Gabor texture analysis. The results showed that the success rate of algorithm of RUS-CHN standard segmenting 13 ossification centers of 980 random samples was 99. 2%.
作者
曹润
熊开宇
黄伟航
姚乐辉
何辉
贺莹莹
CAO Run;XIONG Kai-yu;HUANG Wei-hang;YAO Le-hui;HE Hui;HE Ying-ying(Beijing Sport University, Beijing 100084, China;Physical Education Department, Xuehang University, Xuehang 461000, Henan China)
出处
《北京体育大学学报》
CSSCI
北大核心
2018年第4期75-81,共7页
Journal of Beijing Sport University
基金
中央高校基本科研业务费专项资金资助项目(编号:2016QN013)
北京高校青年英才计划(编号:YETP1258)
关键词
骨龄
骨化中心
图像分割
bone age
ossification center
image segmentation
