基于CT图像的肝脏血管树三维拓扑模型的构建及应用

发布时间：2018-04-21 00:43

本文选题：肝脏血管系统 + 细化　；参考：《重庆大学》2013年硕士论文

【摘要】：对肝脏影像进行三维重建，建立数字化的肝脏模型能够弥补二维影像评估的不足。结构化的血管描述了血管的解剖位置和级数，也体现了供血关系。形态学里的对象的中心线，是一种经过降维的物体形态的描述方式，不但把对象的轮廓和区域信息进行了组合，反映出对象重要的视觉上的线索；并且在将中心线的线形连通结构转化为树或图的抽象形式后，可以对对象进行特征匹配，因而基于中心线的目标表示和识别技术成为模式识别和计算机视觉的重要研究内容。该领域的核心技术是中心线提取技术和基于中心线的目标表示技术。对于前者，已有大量的中心线提取算法被提出，而目前对于后者的研究还很有限，因为直接从中心线图像中提取目标对象的结构特征是困难和低效的。本文提出了一种血管树拓扑结构的图表示方法。首先通过模板匹配对提取出的肝脏血管树进行细化和单体素化，通过分析体素点的邻接关系对分叉点进行标记，提出了一种基于三维连通域标记的广度优先搜索算法来去除环，并利用三维图像中血管的管径和长度信息进行剪枝，提取出符合肝脏血管树实际情况的中心线。然后在此基础上遍历该中心线，同时构造多叉树，得到血管树拓扑结构的图表示。统计结果表明，该方法提取得到的肝脏血管树中心线连通性较好，精确性较高，能够应用到血管分级和分支血管的长度以及管径的测量中去。经过多次实验，利用字典树的数据结构构建的图表示能够很好的抽象表示血管树。肝脏门静脉和肝静脉的管径作为肝病诊断的依据，具有重要意义。本文利用血管树的拓扑结构模型对血管树中心线和血管树进行了划分，然后采用精确的欧氏距离计算血管分支的长度和管径。最后采用Strahler分级方法对血管树进行了分级，以满足医师在诊断时使用不同过滤参数的需要，分级的结果较好的反映了血管树的层级关系。通过Bland-Altman分析，证明本文方法在计算血管的长度和管径时有很高的精确性。该方法在虚拟肝脏手术规划系统中得到了很好的应用，，可在此基础上对肝脏进行分段，辅助外科医师制定手术预案。
[Abstract]:Three-dimensional reconstruction of liver image and establishment of a digital liver model can make up for the deficiency of two-dimensional image evaluation. Structured blood vessels describe the anatomical location and progression of blood vessels, as well as the relationship between blood supply. The central line of the object in morphology is a way of describing the shape of the object after dimensionality reduction. It not only combines the outline of the object with the regional information, but also reflects the important visual clues of the object. After transforming the linearly connected structure of the center line into the abstract form of tree or graph, the object feature matching can be carried out, so the target representation and recognition technology based on the center line has become an important research content of pattern recognition and computer vision. The core technologies in this field are centerline extraction and centerline based target representation. For the former, a large number of centerline extraction algorithms have been proposed, but the research on the latter is still very limited, because it is difficult and inefficient to extract the structural features of the target object directly from the centerline image. In this paper, a graph representation method of vascular tree topology is proposed. Firstly, the extracted hepatic vascular tree is refined and monomeric by template matching, and the bifurcation points are marked by analyzing the adjacency of voxel points, and a breadth-first search algorithm based on 3D connected domain markers is proposed to remove the ring. The information of diameter and length of blood vessel in 3D image is used to prune and extract the centerline which accords with the actual situation of hepatic vascular tree. Then the center line is traversed and the multi-tree is constructed and the graph representation of the topological structure of the vascular tree is obtained. The statistical results show that the proposed method has good connectivity and high accuracy, and can be applied to the classification of blood vessels and the measurement of the length and diameter of branch vessels. After many experiments, the graph representation constructed by the data structure of dictionary tree can represent vascular tree abstractly. The diameter of hepatic portal vein and hepatic vein is the basis for the diagnosis of liver disease. In this paper, the center line of vascular tree and vascular tree are divided by using the topological structure model of vascular tree, and the length and diameter of vascular branches are calculated by using accurate Euclidean distance. Finally, the vascular tree was classified by Strahler classification method to meet the needs of different filtering parameters used by doctors in diagnosis. The classification results reflected the hierarchical relationship of vascular tree. It is proved by Bland-Altman analysis that the proposed method is accurate in calculating the length and diameter of blood vessels. This method has been applied well in the virtual liver surgery planning system. On this basis, the liver can be segmented to assist the surgeon to make the operation plan.
【学位授予单位】：重庆大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TP391.41;R816.5

【参考文献】