基于CT影像的肺气管道骨架化的研究

发布时间：2018-06-06 23:10

本文选题：肺气管道 + 细化　；参考：《东北大学》2012年硕士论文

【摘要】：在全球范围内,呼吸系统疾病是一类重要的非传染性流行疾病。多探测器计算机断层成像(MDCT)是一种有效的、非创伤性气道疾病研究和评价手段。但由于缺乏气管结构分析方法研究,在揭示疾病的本质、特征和发生、发展规律时仍面临挑战。在此背景下,计算机辅助诊断(computer aided diagnosis, CAD)的研究蓬勃发展起来。肺气管道的自动分割与骨架化以及对肺气管相关疾病的自动诊断系统是当前CAD系统的一个重要分支。本课题的目的是提出并实现一种自动的气管骨架化算法,从而提取气道中心线,生成气管树状结构图,获取气管2D横截面,为气管结构的定量化分析提供方法。研究中的CT影像数据均来自于中国医科大学附属盛京医院(中国,沈阳),飞利浦Brillance64CT系统,重建矩阵512*512。基于由CT影像数据中提取的二值化气道树,提出拓扑细化算法,通过简单点判断不断从边界上删除多余体素,直至剩下单体素宽的骨架。再设定合适阈值,将骨架线上的细化“毛刺”剪除。从最终骨架中识别气道根点、分叉点、线端点和线上点,利用二叉树拓扑结构建立气管树状结构图,实现气道级数标记。同时利用提取的气道根点、分叉点和线端点建立简化的树状模型。最后,利用三次B样条曲线对气道中心线平滑后,计算中心线切向量(即气道横截面法向量),从而通过坐标变换获取2D横断面图像数据,为将来气管直径测量和气管壁厚度分析打下基础。本课题提出的算法能成功提取10组数据中的气道骨架,剪枝阈值设置为15较为合适。获得的支气管树结构图和简化树状模型可实现自动命名,最高可标记至12级支气管,总支气管分支数量最多可达174个。最后,快速准确提取各级支气管的2D横断面图像数据。综上所述,本课题提出的自动气管骨架化算法可以用于气管的定量化结构分析,为揭示气道疾病的本质、特征和发生、发展规律提供帮助。
[Abstract]:Globally, respiratory diseases are an important class of non-communicable endemic diseases. Multi-detector computed tomography (MDCT) is an effective and non-traumatic method for the study and evaluation of airway diseases. However, due to the lack of tracheal structure analysis method, it is still facing challenges in revealing the nature, characteristics, occurrence and development of the disease. Under this background, the research of computer aided diagnosis (aided diagnosis, CAD) is booming. The automatic segmentation and skeleton of the pulmonary duct and the automatic diagnosis system for the diseases related to the pulmonary duct are an important branch of the current CAD system. The purpose of this paper is to propose and implement an automatic trachea skeleton algorithm, so as to extract the central line of the airway, generate the trachea tree structure diagram, obtain the trachea 2D cross section, and provide a method for the quantitative analysis of the trachea structure. The CT image data in the study were obtained from Shengjing Hospital affiliated to China Medical University (Shenyang, China, Philips Brillance64CT system, reconstruction matrix 512 / 512). Based on the binary airway tree extracted from CT image data, a topological thinning algorithm is proposed. The redundant voxels are continuously deleted from the boundary by simple point judgment until the skeleton with a single element width is left. Then set the appropriate threshold, the skeleton line on the fine "burr" cut off. The airway root points, bifurcation points, line endpoints and line points are identified from the final skeleton, and the trachea tree structure diagram is established by using the binary tree topology to mark the airway series. At the same time, a simplified tree model is established using the extracted airway root points, bifurcation points and line endpoints. Finally, after the cubic B-spline curve is used to smooth the central line of the airway, the tangent vector of the center line (i.e. the normal vector of the airway cross-section) is calculated, and the 2D cross-sectional image data are obtained by coordinate transformation. It will lay a foundation for the measurement of trachea diameter and the analysis of the wall thickness of trachea in the future. The proposed algorithm can extract the airway skeleton from 10 groups of data successfully and set the pruning threshold to 15. The obtained bronchotree structure diagram and simplified tree model can be automatically named. The highest number of bronchi can be labeled to grade 12, and the number of branches of the total bronchus can be up to 174. Finally, 2D cross-sectional image data of all levels of bronchus were extracted quickly and accurately. To sum up, the automatic trachea skeleton algorithm proposed in this paper can be used to analyze the quantitative structure of trachea, which can help to reveal the nature, characteristics, occurrence and development 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】：2012
【分类号】：R816.41;TP391.41

【参考文献】