肺部气道树骨架的自动提

发布时间：2019-03-19 20:41

【摘要】：人体肺部气道树主要包括气管、主支气管及叶、段各级支气管(约23级分支),是呼吸系统的重要组成部分,其结构和功能的改变是呼吸系统疾病的重要原因和表现。MDCT(Multi-detector computed tomography,MDCT)是一种重要的、非创伤性气道疾病评价手段。利用MDCT,一次屏气即可获得全肺连续和(或)重叠的、近似各向同性的高分辨率薄层(0.75-1.5 mm)结构图像。然而,MDCT获取的海量数据对科学诊断和研究提出了巨大的挑战,需要利用先进影像处理科学实现气道自动提取和结构分析。本文的目的是提出一种适用于肺部气道树结构分析的方法,实现肺部气道树骨架的自动提取、标记和定量化分析。本文使用自己建立的两个模型和26组CT影像数据(均来自于中国医科大学附属盛京医院)进行算法验证。首先,利用建立的模型对拓扑细化算法进行了验证,并提出了分组连通性检验的改进方法。然后,将改进后的细化算法应用于从CT图像中分割提出的气道树模型。具体地,在对图像经过形态学处理后,利用拓扑细化方法提取肺部气管骨架,对肺部气管影像中的体素点进行简单点判断,利用欧拉示性数和连通性两个条件,将简单点删除,得到肺部气道树的骨架。在肺部气道树骨架的基础上,提取骨架的分叉点和叶子节点,对不同级数的支气管使用不同颜色进行连线,得到肺部气道树的树状结构。最后,在树状结构的基础上进行支气管长度和分叉角度的测量。改进后的拓扑细化算法对建立的两个模型可以得到完成正确的细化结果。另外,本文对26组数据进行了实验分析,成功提取所有数据的肺部气道树骨架,并且成功生成了 22组骨架的树状结构,其中树状结构中没有错误分叉的骨架有14组。本文提取的肺部气道树树状结构的级数最高可达到15级,气道树的叶子节点最多有52个。最后,本文对所有数据中主支气管进行的量化分析发现,左主支气管的长度平均值为76.43 mm,右主支气管长度平均值为37.06 mm,左右主支气管之间夹角的平均值为109.76度。所得左右主气管长度和分叉角度与已有文献报道一致。结果表明,本文提出的肺部气道树结构分析的方法,对于部分肺部气道树可以实现骨架的自动提取、标记和定量化分析。肺部气道树的预处理程度对骨架提取至关重要,没有空洞且表面光滑的肺部气道树是正确提取骨架的基础。该方法对于气道树解剖结构测量、气道拓扑结构认识和气道疾病定量化诊断都具有一定的潜在价值。
[Abstract]:The human lung airway tree mainly consists of trachea, main bronchus, lobar and segmental bronchi (about 23 branches), which is an important part of the respiratory system. The structural and functional changes are important causes and manifestations of respiratory diseases. MDCT (Multi-detector computed tomography,MDCT) is an important and noninvasive assessment method for airway diseases. Continuous and (or) overlapping, approximately isotropic, high resolution thin slice (0.75 mm) structural images of the whole lung can be obtained by using MDCT, with one breath hold. However, the massive data obtained by MDCT poses a great challenge to scientific diagnosis and research. It is necessary to use advanced image processing science to realize automatic airway extraction and structural analysis. The purpose of this paper is to propose a method suitable for the structure analysis of pulmonary airway tree, and to realize the automatic extraction, marking and quantitative analysis of the pulmonary airway tree skeleton. In this paper, two models and 26 sets of CT image data (both from Shengjing Hospital, affiliated to China Medical University) are used to verify the algorithm. Firstly, the proposed model is used to verify the topology thinning algorithm, and an improved method of packet connectivity test is proposed. Then, the improved thinning algorithm is applied to the gas tree model of CT image segmentation. Specifically, after morphological processing of the image, the method of topological thinning is used to extract the trachea skeleton of the lung, and the voxel points in the image of the lung trachea are simply judged. The simple points are deleted by using two conditions: Euler number and connectivity. Get the skeleton of the lung airway tree. Based on the skeleton of the pulmonary airway tree, the bifurcation point and leaf node of the skeleton were extracted, and the tree-like structure of the pulmonary airway tree was obtained by connecting the bronchi of different order with different colors. Finally, the length and bifurcation angle of bronchus were measured on the basis of tree structure. The improved topology thinning algorithm can get the correct refinement results for the two models. In addition, 26 sets of data were analyzed experimentally, and the lung airway tree skeleton of all the data was successfully extracted, and 22 groups of tree structure were successfully generated. Among them, there were 14 groups of skeleton without error bifurcation in the tree structure. In this paper, the order of the tree-like structure of the lung airway tree is up to 15, and the number of leaf nodes of the tree is up to 52. Finally, a quantitative analysis of the main bronchus in all data shows that the average length of the left main bronchus is 76.43 mm, and the average length of the right main bronchus is 37.06 mm,. The average angle between the left and right main bronchi is 109.76 degrees. The length and bifurcation angle of the left and right main trachea are consistent with those reported in the literature. The results show that the method proposed in this paper can automatically extract, label and quantify the skeleton of some of the lung airway trees. The preprocessing degree of the lung airway tree is very important to the skeleton extraction. There is no cavity and smooth surface of the lung air channel tree is the basis for the correct extraction of the skeleton. This method is of potential value for the measurement of anatomical structure of airway tree, the recognition of airway topology and the quantitative diagnosis of airway diseases.
【学位授予单位】：东北大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：R816.4;R56

【相似文献】