病变粘连髋关节目标分割算法  

Target segmentation algorithm for diseased adhesive hip joint

作　　者：王永琪 张远民 王学渊[1] WANG Yongqi;ZHANG Yuanmin;WANG Xueyuan(School of Information Engineering,Southwest University of Science and Technology,Mianyang 621000,China;Unicom(Sichuan)Industrial Internet Research Institute,Chengdu 610000,China)

机构地区：[1]西南科技大学信息工程学院,四川绵阳621000 [2]联通(四川)产业互联网研究院,四川成都610000

出　　处：《传感器与微系统》2022年第9期113-115,120,共4页Transducer and Microsystem Technologies

基　　金：国家自然科学基金资助项目(61771411)。

摘　　要：针对病变髋关节灰度分布不均匀图像粘连造成的分割难题,提出了一种改进Otsu双阈值和分水岭拟合圆的分割算法。首先对图像使用形态学增强与改进的Otsu双阈值算法得到粘连髋关节的二值图像,再利用分水岭结合拟合圆方法,完成病变髋关节的分割。本实验数据为20例病变髋关节,通过与医学处理软件ITK-Snap、分水岭算法对比,本文方法的平均分割精度为97.51%。实验结果表明:该方法不仅能够提高髋关节的分割精度,同时还可以避免不同医生手动分割造成的主观差异。A segmentation algorithm is proposed, aiming at the problem caused by adhesion of image of uniform distribution of gray scale of diseased hip joint.An improved Otsu dual-threshold and circle fitting watershed segmentation algorithm is proposed to segment the femur head from the acetabulum.The method uses morphological enhancement and the improved Otsu dual-thresholding segmentation to obtain the binary image of the adhesive hip joint, and then uses the circle fitting watershed method to segment the femur head from the acetabulum.20 diseased patient data are compared with ITK-Snap and watershed segmentation method, the average segmentation precision of proposed algorithm is 97.51 %.The results show that the proposed algorithm can not only improve segment precision of hip joint, but also avoid subjective differences caused by manual segmentation.

关 键 词：髋关节分割 双阈值 带标记分水岭 拟合圆 病变髋关节 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]