基于三维区域生长法的CT图像肺部血管分割

发布时间：2018-06-28 03:46

本文选题：肺部血管分割 + 三维区域生长算法　；参考：《山东师范大学》2017年硕士论文

【摘要】：计算机断层扫描CT影像具有空间分辨率高等特点,已成为肺部疾病影像学诊断的首选方式,而医生多数采用肺部强化CT图像来对肺部疾病诊断和治疗。CT图像中肺部血管的分割,是肺炎、肺动脉栓塞等肺部疾病计算机辅助检测和诊断的基础与关键,具有重要的研究意义与应用价值。在文献阅读的基础上,结合肺部血管的特点,提出一种基于三维区域生长法的肺部血管分割算法,主要有三方面的研究工作。(1)相关理论基础的学习与研究。学习了肺部相关医学和医学图像知识,以及肺部血管和气管的走行。研究了现阶段肺部血管的分割方法,并且重点分析了基于区域生长算法肺部血管分割近几年的研究成果。通过阅读文献,明晰了现阶段国内外关于肺部血管分割的研究现状。(2)算法设计与分析。整个算法主要包括图像预处理、肺泡分割、肺气管分割、肺血管分割等4个方面。针对肺泡分割,提出漫水填充分割算法,解决了左右肺连接和过分分割问题;针对肺部气管分割,提出了一种26-邻域三维区域生长算法,精确分割出肺部气管;针对肺血管分割,首先将肺气管像素从CT图像中剔除,随后应用26-邻域三维区域生长法分割。(3)算法实现与验证。利用软件平台Visual Studio 2010,结合Open CV和VTK等第三方开源库,运用C++语言编程实现算法,通过实验结果来验证算法的可行性与有效性,运用图像分割评价指标给出定量评价。研究的创新之处是,(1)使用双阈值作为三维区域生长的生长规则,在种子点手动选取之后,根据方差和灰度均值作为生长规则,从而提高分割的精度。(2)针对肺泡分割存在易过度分割、左右肺部不能分离等问题,提出一种全自动的基于漫水填充算法的肺泡分割方法。研究存在的不足是,肺毛细血管分割不完全,分割的自动化程度需进一步提高。
[Abstract]:Computed tomography (CT) has the characteristics of high spatial resolution and has become the first choice for imaging diagnosis of lung diseases. Most doctors use enhanced CT images of lung to diagnose lung diseases and treat pulmonary vessels in.CT images, which are the basis of computer aided detection and diagnosis of pneumonia, pulmonary embolism and other pulmonary diseases. Base and key, it has important research significance and application value. On the basis of literature reading, combined with the characteristics of pulmonary blood vessels, a new algorithm of pulmonary vascular segmentation based on three dimensional region growth method is proposed, which mainly has three aspects of research. (1) study and Research on the basis of related theories. Knowledge, and the movement of pulmonary vessels and trachea. The segmentation methods of pulmonary vessels at the present stage are studied, and the results of recent years' research on the segmentation of pulmonary vessels based on regional growth algorithm are emphatically analyzed. Through reading literature, the current research status on pulmonary vascular segmentation at home and abroad is clarified. (2) algorithm design and analysis. The whole algorithm is designed and analyzed. Mainly including image preprocessing, alveolar segmentation, lung trachea segmentation, pulmonary vascular segmentation, and other 4 aspects. Aiming at alveolar segmentation, a diffuse filling segmentation algorithm is proposed to solve the problem of left and right lung connections and excessive segmentation. A new 26- neighborhood three-dimensional region growth algorithm is proposed for the segmentation of lung and trachea, and the lung trachea is segmented accurately. Segmentation, first remove the lung trachea pixels from the CT image, then use the 26- neighborhood three-dimensional region growth method. (3) the algorithm implementation and verification. Using the software platform Visual Studio 2010, combined with Open CV and VTK, the three party open source library, using C++ language programming to implement the algorithm, through the experimental results to verify the feasibility and effectiveness of the algorithm. Quantitative evaluation is given with the evaluation index of image segmentation. The innovation of the study is: (1) using double threshold as the growth rule of three-dimensional region growth, after the manual selection of the seed points, according to the variance and the mean value as the growth rule, and thus improve the precision of the segmentation. (2) there is an easy over segmentation for the pulmonary alveolus segmentation, and the left and right lungs can not be divided. In order to solve the problem, a fully automatic method of pulmonary alveolus segmentation based on diffuse filling algorithm is proposed. The shortcomings of the study are that the segmentation of lung capillaries is incomplete and the degree of automation of the segmentation needs to be further improved.
【学位授予单位】：山东师范大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：R816.4;TP391.41

【参考文献】