基于Hessian矩阵和GMM-EM算法的肝脏三维血管树提取  

3D Liver Vessel Segmentation Based on Hessian Matrix and GMM-EM Algorithm

作　　者：皮净锐[1] 房斌[1] 王翊[1] 刘润宗[1] 

机构地区：[1]重庆大学计算机学院,重庆400030

出　　处：《生物医学工程学杂志》2013年第3期486-492,共7页Journal of Biomedical Engineering

摘　　要：血管系统的准确分割是许多医学应用的基础。本文提出了一种基于Hessian矩阵和高斯混合模型的最大期望(GMM-EM)算法提取肝脏三维血管树的有效方法。首先采用Hessian矩阵对肝脏原始图像进行管状物探测和增强,接着采用GMM-EM分割算法得到粗略血管系统;由于Hessian矩阵对噪声敏感,容易出现血管的断落,本文采用形态学闭操作减少断落,同时设计了空间滤波器解决血管尾影问题;最后利用三维空间连通域搜索算法去除噪声,实现血管系统的空域连通。实验表明,本文的方法能有效提取肝脏的血管树。An accurate segmentation of vascular systems is fundamental for many medical applications. In this paper, we propose a 3D vessel enhancement and extraction method. It is based on the analysis of Hessian matrix and Gaussi- an mixture model-expectation-maximization （GMM-EM） algorithm. Firstly, tube-like vessels were detected and en- hanced based on the Hessian matrix eigenvalues. And then, the vascular system was segmented, and then a rough system was obtained with GMM-EM. Hessian-based filters were found to be sensitive to noise and sometimes gave discontinued vessels. Hence, we utilized the closing operation to avoid discontinuity and a 3D-filter on the segmented vessels to reduce noise brought by the contrast agent. Finally, a searching method based on spatial connected area is presented to connect the vascular system in 3D. The experimental results illustrated the efficiency of the method for 3D liver vessel segmentation proposed in this paper.

关 键 词：肝脏血管系统 HESSIAN矩阵 高斯混合模型的最大期望算法 三维连通血管树 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]