冠脉CT造影图像的血管分割与斑块祛除

发布时间：2018-04-05 03:09

本文选题：图像分割　切入点：钙化斑块　出处：《北京邮电大学》2017年硕士论文

【摘要】：冠心病的死亡率一直居高不下,是一种全世界范围内普遍存在的疾病。为了能够对冠心病患者血管的狭窄程度进行精确诊断,医疗工作者采用了目前最具有临床意义的指标---血流储备分数(Fractional Flow Reserve,FFR)。然而,计算FFR之前需要得到患者冠状动脉血管三维重建的精确模型,这就涉及到本文的两个主要研究点---血管分割以及钙化斑块的检测及祛除。传统的图像分割算法大体上可以归纳为两类:基于边缘的方法和基于区域的方法。前者通过检测相邻像素特征值的突变来获得不同区域之间的边界,再把边界连接起来,从而将图像分成不同的区域,如各种微分算子边缘检测法。后者则是按照事先定义的准则,将某些局部特征(如灰度、纹理等)相似的像素或小区域合并为更大的区域,如区域生长算法,分水岭算法等。心脏区域的血管纵横交错,分布极其复杂,再加上其在CT图像中像素值差异不明显,传统的图像分割算法就显得力不从心了。本文分析了传统图像分割算法的优缺点并结合冠状动脉血管的走行方向以及各分支的粗细、长短等特点,提出了一种适用于冠状动脉血管的三维图像分割算法。实验结果表明,该算法分割精度更高,并且效率也有很大提升。除此之外,患者冠脉血管壁上附着生长的钙化斑块会影响血流速度,如果在CT图像中不加以祛除,会对FFR计算结果的精度产生严重影响,所以,需要将CT图像中的钙化斑块检测出来并祛除。为了降低斑块检测的误检率,本文采用了 HOG特征和LBP特征结合SVM分类器的方法对钙化斑块进行了初步的检测。接下来,在前人研究成果的基础上,提出了定向膨胀算法和基于血管中心线的斑块轮廓拟合算法,实现斑块祛除。经过这两步的处理,完成了钙化斑块的检测及祛除。通过对实验结果的分析比较,证明了本文的方法在斑块的检测及祛除方面达到了比较高的精确度。
[Abstract]:Coronary heart disease mortality has been high, is a worldwide widespread disease.In order to accurately diagnose the degree of vascular stenosis in patients with coronary heart disease (CHD), medical workers have adopted Fractional Flow Reserve fraction (Fractional Flow Reserve), which is the most important clinical marker.However, it is necessary to obtain an accurate model of three-dimensional reconstruction of coronary artery before calculating FFR, which involves two main research points in this paper-blood vessel segmentation, detection and removal of calcified plaque.Traditional image segmentation algorithms can be classified into two categories: edge-based method and region-based method.In the former, the boundary between different regions is obtained by detecting the mutation of adjacent pixel eigenvalues, and then the boundary is connected together, thus the image is divided into different regions, such as the edge detection methods of various differential operators.On the other hand, according to the pre-defined criteria, some similar pixels or small regions with similar local features (such as grayscale, texture, etc.) are merged into larger regions, such as region growth algorithm, watershed algorithm and so on.The blood vessels in the heart region are crisscrossed and the distribution is extremely complex. Besides, the pixel value difference in CT images is not obvious, so the traditional image segmentation algorithm is not strong enough.In this paper, the advantages and disadvantages of traditional image segmentation algorithms are analyzed, and a 3D image segmentation algorithm suitable for coronary artery segmentation is proposed, which combines the characteristics of the direction of coronary artery and the thickness and length of each branch.Experimental results show that the algorithm has higher segmentation accuracy and higher efficiency.In addition, calcified plaques attached to the coronary artery wall affect blood flow velocity, and if not removed from CT images, the accuracy of the FFR results can be seriously affected, so,Calcified plaques in CT images need to be detected and removed.In order to reduce the false detection rate of plaque detection, the method of HOG feature and LBP feature combined with SVM classifier is used to detect calcified plaque.Then, on the basis of the previous research results, the directional expansion algorithm and the patch contour fitting algorithm based on the center line of blood vessels are proposed to achieve plaque removal.After these two steps of treatment, the detection and removal of calcified plaques were completed.Through the analysis and comparison of the experimental results, it is proved that the proposed method achieves high accuracy in plaque detection and removal.
【学位授予单位】：北京邮电大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：R816.2;TP391.41

【相似文献】