肝脏血管中心线及分级数据的提取方法

发布时间：2018-11-23 20:28

【摘要】：随着肝脏部分切除术特别是活体肝移植的发展,肝脏分段解剖在手术中越来越重要。中心线及分级数据作为血管的结构化描述方法,不仅是肝脏分段的重要依据,而且描述了血管的解剖位置和级数,也体现了供血关系,因此准确提取肝脏血管的中心线及分级数据是肝脏临床诊断和分析过程中一个重要环节。由于肝脏血管分支复杂,并且肝脏血管的分割结果不平滑,使用传统方法提取血管中心线和分级数据提取方法时,存在错误分支多、血管分叉处像素冗余和中心线不连续等缺陷。在提取中心线和分级数据过程中通过剪枝策略去除细小分支,并且引入一种冗余像素判断方法去除中心线中的多余像素就能提高传统方法的鲁棒性,最终提取准确的血管中心线和分级数据。中心线提取算法的基础是快速步进法(FMM),计算过程中根据FMM速度函数的不同,计算不同速度的场。首先设定FMM速度函数为常函数,计算血管中每一个像素到边缘的最短距离,获得低速场,自动选取距离场中值最大的像素作为血管源点Ps。然后设定FMM的速度函数是以低速场值为自变量的开方函数,计算血管每一个像素点到源点Ps的最短距离,获得中速场,将中速场取整分块,通过宽度优先遍历的方法获得血管端点。最后设定FMM的速度函数是以距离场值为自变量的线性函数,计算血管每一个像素到源点Ps的最短距离,获得高速场,在高速场中利用梯度下降法从血管端点回溯到源点Ps,获得血管中心线。在提取分级数据的过程中,首先根据中心线上点的几何特征,标记出分叉点,然后从血管主干点开始遍历中心线,构建血管结构字典树,为了准确表示血管的结构特征,利用修改局部点的方法去除字典树中的噪声节点,合并相邻的分支节点,最后根据血管特征提取准确的分级数据。实验证明,对于大部分病例数据,以上可以快速、有效地提取血管中心线和分级数据。
[Abstract]:With the development of partial hepatectomy, especially in vivo liver transplantation, segmental anatomy of liver is becoming more and more important. Centerline and grading data are not only the important basis of liver segmentation, but also describe the anatomical position and progression of blood vessels, and reflect the relationship between blood supply and blood supply. Therefore, accurate extraction of liver blood vessel centerline and grading data is an important link in the process of liver clinical diagnosis and analysis. Because the branches of hepatic blood vessels are complex and the segmentation results of hepatic blood vessels are not smooth, there are many wrong branches when traditional methods are used to extract blood vessel centerline and classification data. Defects such as pixel redundancy and discontinuity of the center line at the vascular bifurcation. In the process of extracting centerline and grading data, the small branches are removed by pruning strategy, and the robustness of traditional methods is improved by introducing a redundant pixel judgment method to remove redundant pixels from the centerline. Finally, accurate blood vessel centerline and grading data were extracted. The basis of centerline extraction algorithm is to calculate the field of different velocities according to the difference of FMM velocity function in the process of (FMM), calculation by fast step method. First, the FMM velocity function is set as a constant function, the shortest distance from each pixel to the edge of the vessel is calculated, and the low velocity field is obtained. The pixel with the largest median distance field is automatically selected as the vascular source point Ps.. Then the velocity function of FMM is an open-square function with the low velocity field value as the independent variable. The shortest distance from each pixel point to the source point Ps is calculated, the medium velocity field is obtained, and the intermediate velocity field is divided into blocks. The vascular endpoints are obtained by width-first traversal. Finally, the velocity function of FMM is a linear function with the distance field value as the independent variable. The shortest distance from each pixel to the source point Ps is calculated, and the high-speed field is obtained. In the high-speed field, the gradient descent method is used to trace back to the source point Ps, from the vascular endpoint. The central line of the blood vessel was obtained. In the process of extracting the grading data, the bifurcation points are first marked according to the geometric characteristics of the points on the center line, and then the center line is traversed from the main point of the blood vessel to construct a dictionary tree of the blood vessel structure, in order to accurately represent the structural characteristics of the blood vessel. The noise nodes in the dictionary tree are removed by modifying the local points, the adjacent branch nodes are merged, and the accurate classification data are extracted according to the vascular features. Experimental results show that the above method can extract the blood vessel centerline and grading data quickly and effectively for most case data.
【学位授予单位】：华中科技大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：R657.3;TP391.41

【参考文献】