摘要
基于计算机断层血管造影(Computed Tomography Angiography,CTA)的冠状动脉自动分割是后续冠脉狭窄和斑块等病灶识别的重要前置步骤。算法首先结合窗宽窗位调整和基于灰狼优化算法的多级阈值处理,实现了对肺部血管的抑制,提升了冠状动脉和背景组织的对比度。然后基于升主动脉和冠脉的三维解剖结构特征,采用光流法识别升主动脉根部的冠状动脉起始层,为后续冠状动脉分割提供起始种子点。最后利用结合端点检测的自适应区域生长法提取完整的冠状动脉。在20例CTA数据上的实验结果表明,相比于不带端点检测的区域生长算法,本文算法的Dice和Jaccard系数分别提高了4%和8%,达到了0.70和0.57,MSD和MAXSD分别降低了2%和4%,达到了0.40和2.66。所提算法实现了冠状动脉树的自动提取,减少了人工干预,克服了传统区域生长法易产生的过分割现象以及对细小血管的漏分割现象,提高了冠状动脉分割的准确性。
Automatic coronary segmentation based on computed tomography angiography(CTA)is an important step before coronary lesion detection.The proposed algorithm first combines window width and window level adjustment and multi-level threshold processing based on the gray wolf optimization algorithm to achieve suppression of pulmonary blood vessels and promote the contrast of coronary arteries and background tissue.Then,based on the three-dimensional anatomical features of ascending aorta and coronary artery,optical flow method is used to identify the starting layer of coronary artery at the root of ascending aorta to provide starting seed points for subsequent coronary segmentation.Finally,the branch based adaptive region growing algorithm(BARG)is used to extract the complete coronary artery.The results on 20 CTA data show that,compared with the region growing algorithm without branch detection,the Dice and Jaccard coefficients of the proposed method are increased by 4%and 8%,reaching 0.70 and 0.57,and the MSD and MAXSD are reduced by 2%and 4%respectively,reaching 0.4 and 2.66.The proposed algorithm realizes the automatic extraction of coronary artery tree,reduces manual intervention,overcomes the over-segmentation phenomenon and the omission segmentation phenomenon of small blood vessels easily produced by the traditional region growing method,and improves the accuracy of coronary artery segmentation.
作者
罗宇杰
祝磊
马骏
薛凌云
张子恒
徐平
刘亦安
严明
LUO Yujie;ZHU Lei;MA Jun;XUE Lingyun;ZHANG Ziheng;XU Ping;LIU Yian;YAN Ming(School of Automation,Hangzhou Dianzi University,Hangzhou Zhejiang 310018,China;Information Center,Zhejiang Provincial People’s Hospital(Affiliate People’s Hospital of Hangzhou Medical College),Hangzhou Zhejiang 310014,China)
出处
《传感技术学报》
CAS
CSCD
北大核心
2022年第11期1491-1498,共8页
Chinese Journal of Sensors and Actuators
基金
国家自然科学基金——浙江两化融合基金重点项目(U1609218)。
关键词
医学图像处理
冠状动脉分割
灰狼优化算法
光流法
区域生长法
medical image processing
coronary artery segmentation
gray wolf optimization
optical flow
region growing
