【作者】 徐智；

【导师】 郁道银；

【作者基本信息】 天津大学 ， 光学工程， 2003， 博士

【摘要】 心血管造影图像已经成为心血管疾病诊断的金标准。利用多角度二维心血管造影图像重建的三维心血管图像具有客观、形象的特点，能够实现病变的量化描述，对心血管狭窄的预防、诊断和治疗具有重要意义。利用造影图像重建三维心血管包括两个阶段：1、二维心血管信息提取；2、血管三维结构重建。本文工作是围绕心血管造影图像中的心血管骨架信息提取、直径信息提取、主干血管辅助诊断而展开的，并对单面造影法心血管三维重建进行了初步研究，主要内容如下：1、心血管造影图像的灰度特征分析根据造影成像理论和心血管形状特征深入分析了造影图像，归纳出6个基本灰度特征，如血管骨架垂线上的灰度成高斯函数分布等，为二维信息提取提供了依据。2、血管骨架提取本文首先对造影图像进行了预处理，采用了形态学Top-Hat变换、旋转一维高斯模板等方法对图像进行了噪声去除和图像增强。然后提取血管区域，在对自适应阈值、最大类间方差自动阈值等方法进行大量实验研究后，提出了自适应跟踪圆模板法，该方法经过人工初始化后可以对心血管区域进行自动提取。最后进行骨架提取，通过对比研究直接法和间接法，本文采用了间接法，即通过细化血管区域实现心血管骨架提取。在血管区域细化过程中，提出了连通数细化方法，该方法具有计算量小、运算速度快、各向同性等特征。3、血管连续边缘提取心血管连续边缘提取是三维重建和血管辅助诊断的基础。本文在对比研究了多种常规边缘提取方法以及小波多尺度边缘提取法等基础上，提出了最大成本法。该方法首先利用常规微分边缘提取法建立成本空间，然后对血管边缘进行区域划分，利用动态规划理论提取最大成本曲线作为指定血管的边缘。最大成本法不仅具有常规边缘提取方法的优点，而且其提取的边缘具有连续性、单像素宽等优点。4、血管直径测量与主干血管狭窄辅助诊断心血管直径测量和主干血管狭窄辅助诊断均是建立在心血管骨架和边缘提取的基础上。首先通过骨架对血管二次采样，在采样点处以血管两边缘间的最小距离作为该处直径，采用边缘直线拟合方法消除图像离散性给直径测量带来的误差影响。然后利用本文方法测得的血管直径对血管的狭窄进行了定量诊断，其结果与多位医生目视结果具有良好的一致性。5、心血管三维重建的初步研究本文对心血管三维结构模型，心血管三维重建过程模型，三维重建直径信息提取，血管树状结构描述做了初步研究。提出了采用二叉树结构对心血管图像进行结构描述、数据存贮、信息遍历，为三维重建骨架点匹配奠定了基础。更多还原

【Abstract】 The heart vessel angiogram is the gold tool for diagnosing the presence of coronary artery disease. The traditional methods suffer from inter- and intra-observer variability, so it is very important to develop an objective and precise system to auto-diagnose coronary disease. The image of the 3-D heart vessel reconstructed from 2-D angiograms with different angles naturally and quantitatively depicts the vessel stenosis, and the 3D reconstruction is helpful to prevent and cure the stenosis. The procedure of the 3-D reconstruction is divided into two steps: one is to get the 2-D heart vessel informations from the angiogram; the other is to reconstruct 3D vessel. The main work of this thesis is carried out about the detection of the vessel skeleton and boundary, auxiliary diagnosis and of the stenosis of the main vessel and 3-D reconstruction. The works can be included in five parts as follows:1、Analysis of the intensity characters of the heart vessel angiogramThe heart vessel angiogram is deeply analyzed on the base of the imaging theory and the vascular shape. Six intensity characteristics are acquired, which are the rules for the detection of the 2-D information in the angiogram.2、Detection of the heart vessel skeletonImage pre-processing technologies are firstly used to eliminate noise and enhance the image, such as Top-Hat transform, the rotated 1-D Gauss template and so on. Then the self-adaptive tracking circle template is addressed to extract the vessel area on the research of the self-adaptive threshold, maximum classes square error, et al. After initialized by manual supervision, it can get the vessel area from the angiogram enhanced by the rotated 1-D Gauss template. At last, the un-direct method that thins the vessel area as the skeleton is adopted on the research of the direct and un-direct methods. The connectivity-number thinning method is presented to skeletonize the vessel area. It possesses virtues of small computation, low time-cost, isotropy, and so on. 3、Detection of the continuous vessel boundaryThe continuous vessel boundary is the base for measuring the diameter of the 3-D reconstruction and diagnosing the vessel stenosis. The Max-Cost(MC) is presented on the base of the studies of the traditional edge detecting methods and Wavelet. It is consisted of two steps: the fist step is that the cost field is built up by the traditional edge detecting method—differential method; The second step is that the boundary, the curve with maximal cost in the segmented area, is detected serially by the dynamic programming (DP). The MC inherits the merits of the traditional method, and its boundary is continuous and one-pixel wide. The minimal diameter of the vessels whose boundary can be detected reaches 1mm. 4、The diameter measurement and auxiliary diagnose of the main vesselThe diameter measurement and auxiliary diagnosis are based on the skeleton and <WP=5>continuous boundary of the heart vessel. The vessel is re-sampled by the skeleton, and the minimal distance of the vessel edge of the re-sampled point is defined as the corresponding vessel diameter. The linear fitting is used to weaken the discrete error of the diameter. Then the vessel diameter information detected by the above method is used to diagnose the vessel stenosis, and its result is well consistent with the diagnoses of experienced doctors. 5、Research on the heart vessel 3-D reconstructionThe model of the 3-D heart vessel and 3-D reconstruction, the measurement of the diameter for 3-D reconstruction and description of the vessel tree are briefly studied. This thesis presents that the binary tree is used to descript the vessel tree, store the data and traverse the information. It makes preparation for the matching of vessel skeleton point in the process of 3-D reconstruction and supplies a good data surface.更多还原