摘要
本研究旨在从心脏双源CT数据中自动精确分割出冠状动脉。采用一种基于多尺度滤波和概率决策的血管自动分割算法。先基于多尺度Hessian矩阵增强图像中的管状结构,再利用最大后验概率基于灰度将体素分为目标和背景2类,最后用26邻域区域生长法分割出左冠状动脉。实验结果表明,可精确分割出冠状动脉并提取血管中心线。该算法避免了血管泄露问题,无伪血管,无需人工交互,是一种有效的双源CT冠状动脉自动提取方法。
To solve the accurate coronary artery segmentation,an automatic segmentation algorithm based on multi-scale filtering and statistic decision were proposed. First,the tubular structure based on multi-scale Hessian matrix was enhanced,then the voxels into target and background were divided,at last the left arteries was extracted by the process of 3D region. The experiment results showed that the method could segment coronary artery and extract centerline of the vessels accurately. The proposed model can avoid vascular leakage and pseudo vascular,extract the whole vessel trees automatically,accurately and efficiently without manual interaction.
出处
《生物医学工程研究》
北大核心
2016年第3期197-201,共5页
Journal Of Biomedical Engineering Research
基金
广东省普通高校青年创新人才项目(自然科学)(2015KQNCX071)
广东省科技计划项目(2014A020212084)
关键词
双源CT
冠状动脉
多尺度
HESSIAN矩阵
EM算法
Dual source CT
Coronary artery
Multiple scales
Hessian matrix
Expectation Maximization algorithm