基于概率决策和区域生长的冠状动脉三维分割  

3D Segmentation of Coronary Arteries Based on Probabilistic Decision and Region Growing

作　　者：张朝霞[1] 李宏霄[1] 陈晓冬[1] 郁道银[1] 贾忠伟[2] 浦奎[2] 王星[2] 

机构地区：[1]天津大学精密仪器与光电子工程学院光电信息技术科学教育部重点实验室,天津300072 [2]中国人民解放军第二五四医院,天津300142

出　　处：《天津大学学报》2011年第5期454-457,共4页Journal of Tianjin University(Science and Technology)

摘　　要：冠状动脉的三维分割是多层螺旋CT(MSCT)心血管成像的关键算法之一,良好的分割结果可为医生的诊断提供可靠的依据.运用各向异性扩散滤波技术对冠状动脉的MSCT图像进行了预处理.对滤波后的三维数据采取基于概率决策和区域生长相结合的方法进行冠状动脉的分割.利用最大后验概率决策按照灰度值将体素分为血液、心肌和肺3种区域后,再采用三维区域生长算法提取出重要分支.实验结果表明,算法可以成功地提取出冠状动脉的主要分支,对冠状动脉CT图像的三维可视化重建及冠心病临床诊断有重要意义.3D segmentation of coronary arteries is one of the key algorithms for multi-slice spiral computed tomogra-phy（MSCT）cardiovascular imaging.Satisfactory segmentation results can provide reliable basis for diagnosis.Anisot-ropic diffusion filtering is used to accomplish the pre-processing of MSCT coronary arteries angiography images.Then a hybrid segmentation of coronary artery is carried out,which combines the maximum a posteriori（MAP） probabil-istic decision classification and region growing technology.Voxels are classified into three parts：blood,myocar-dium and lung with MAP Probabilistic Decision classification,followed by the process of 3D region growing to ex-tract the important arteries.Experimental results demonstrate the success of the algorithm to segment the coronary arteries,and therefore it contributes a lot to the 3D visualization reconstruction of the coronary arteries angiography and clinical diagnosis.

关 键 词：冠状动脉 多层螺旋CT 三维分割 各向异性扩散 概率决策 区域生长 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]