基于MSCT影像的气管树分割算法设计与实现

发布时间：2018-03-24 08:06

本文选题：气管树分割　切入点：区域生长　出处：《东北大学》2013年硕士论文

【摘要】：慢性阻塞性肺病(慢阻肺)是发病数量最多的呼吸系统疾病之一,不仅严重影响患者身体健康和生活质量,而且给患者家庭和社会造成很大经济负担。随着多层螺旋CT (MSCT)技术的出现和发展,基于MSCT影像的定量评估可检测肺部组织病变程度,对发病机理进行长期研究,跟踪评估治疗效果,成为慢阻肺诊断和治疗的有力工具。气管树分割是基于MSCT影像的慢阻肺定量评估系统中的关键技术,从大量MSCT影像中实时准确地获取气管树数据已成为当前迫切需要解决的难题。在上述背景下,本文以MSCT图像为数据源,肺部气管树为主要研究对象,对肺功能分析辅助诊断中涉及到的气管树分割与中心线提取算法进行了设计与实现。首先,本文对课题所涉及的气管树解剖学与图像学先验知识、医学图像分割、中心线提取、运动目标跟踪等背景知识进行介绍；其次,设计并实现了基于最大类间方差和多尺度三维区域生长的气管树自动分割算法,提取气管树主干数据；再次,根据数字拓扑学原理设计并实现了基于拓扑细化的气管树中心线提取算法,引用SWC格式标准获得中心线拓扑结构,利用最小二乘法拟合中心线分支方向；最后,将运动目标跟踪技术应用于气管树分割领域,在获得数据基础上设计并实现了基于Kalman滤波的气管树末梢分割算法。实验表明,本文提出的气管树分割算法具有良好精度与速度,满足慢阻肺临床辅助诊断需求。
[Abstract]:Chronic obstructive pulmonary disease (COPD) is one of the most common respiratory diseases, which not only seriously affects the health and quality of life of patients. With the emergence and development of multi-slice spiral CT (MSCT) technology, quantitative assessment based on MSCT images can detect the degree of lung tissue lesions and study the pathogenesis of the disease. Tracheal tree segmentation is a key technique in quantitative assessment system of chronic obstructive pulmonary disease (COPD) based on MSCT image. Obtaining trachea tree data in real time and accurately from a large number of MSCT images has become an urgent problem. Under the above background, this paper takes MSCT image as data source and lung trachea tree as main research object. The trachea tree segmentation and centerline extraction algorithms are designed and implemented in lung function analysis assisted diagnosis. Firstly, the prior knowledge of trachea tree anatomy and imageology, medical image segmentation and centerline extraction are discussed in this paper. Background knowledge such as moving target tracking is introduced. Secondly, an automatic trachea tree segmentation algorithm based on maximum inter-class variance and multi-scale three-dimensional region growth is designed and implemented to extract trachea tree trunk data. According to the principle of digital topology, a trachea tree centerline extraction algorithm based on topology thinning is designed and implemented. The center line topology structure is obtained by using SWC format standard, and the center line branch direction is fitted by the least square method. The trachea tree segmentation algorithm based on Kalman filter is designed and implemented by applying the moving target tracking technology to trachea tree segmentation. Experiments show that the proposed trachea tree segmentation algorithm has good accuracy and speed. To meet the needs of clinical diagnosis of COPD.
【学位授予单位】：东北大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：R563.9;TP391.41

【参考文献】