基于MR图像和三维超声心动图的左心室运动跟踪研究

发布时间：2018-10-08 10:26

【摘要】：据统计，，心脏疾病已经逐渐成为了当今社会人类的头号杀手，正严重威胁和危害着人民群众的健康。左心室通过周期性的收缩、舒张、扭转等运动形式将左心室内的血液泵入全身各组织器官，在心脏活动中的作用尤其重要。因此，可将其运动形式异常、形态异常等作为判断心脏是否产生病变的重要参考依据。研究表明，在很多疾病条件下，心脏运动的程度及形式会产生较大变化。因而，对左心室进行运动跟踪对于心功能正常与否的判定具有积极的意义。实时三维超声技术和心脏MR技术均是无创的心脏运动观察方法，本文将图像处理相关技术，如图像滤波、图像配准、时间序列图像分割以及心室轮廓边缘跟踪等运用于左心室的运动跟踪，针对实时三维超声技术得到的左心室时间序列图像，初步对心脏左心室运动情况进行跟踪分析，并进行了左心室一个心动周期内的三维可视化重建。本文的主要工作以及取得的成果如下： ①进行了左心室超声图像预处理工作。综合几种常见的非线性滤波方法中值滤波、Lee滤波、形态学滤波对左心室超声图像进行滤波处理，通过结果对比分析选取了形态学滤波方法作为图像预处理方法。 ②进行了左心室MR图像配准的研究。针对心脏磁共振图像提出了一种新颖的Radon变换与功率谱结合的图像配准算法。该算法进行的边缘检测过程不仅滤除了部分噪声干扰，还使得输入数据大大减少。利用Radon变换降维、快速提取有用信息的特点和功率谱的旋转不变特性，很好地实现了左心室MR图像配准。 ③提出两种左心室超声图像分割方法。第一种分割方法为利用时刻独立脉冲耦合神经网络针对心脏扭转运动中的左心室壁进行精确定位以及边缘轮廓提取；第二种分割方法利用与投影方法结合C-V水平集分割方法实现了左心室超声序列图像的自动分割，为进一步图像配准提供了优秀的素材。算法首先综合投影方法对左心室位置进行粗定位，然后再对定位区域运用C-V水平集方法进行细分割。 ④实现了左心室运动跟踪以及三维重建。对分割后左心室超声时间序列图像进行了边缘人工标记跟踪并对运动参数进行测量统计分析；在OpenGL平台下，利用最短对角线法对左心室腔轮廓进行三维表面重建，实现了基于轮廓拼接的左心室三维重建。本文提出的基于图像处理的心脏运动跟踪算法可行性高，其中涉及的图像配准、图像分割等算法具有独创性，有一定的临床应用价值，希望在以后的研究中能进一步完善分割以及运动跟踪算法并尽快应用于临床分析。
[Abstract]:According to statistics, heart disease has gradually become the leading killer of human beings, which is seriously threatening and endangering the health of the people. The left ventricle pumps blood into the whole body through periodic contraction, relaxation and torsion, which plays an important role in cardiac activity. Therefore, it can be used as an important reference to judge whether the heart has pathological changes. Studies have shown that the degree and form of cardiac movement vary greatly under many disease conditions. Therefore, the left ventricular movement tracking is of positive significance for the determination of normal cardiac function. Real-time three-dimensional ultrasound and cardiac MR are non-invasive methods to observe cardiac motion. In this paper, image processing techniques, such as image filtering, image registration, etc. Time series image segmentation and contour tracking are used to track the motion of the left ventricle. In view of the time series images of the left ventricle obtained by real-time three-dimensional ultrasound, the motion of the left ventricle is tracked and analyzed preliminarily. Three-dimensional visual reconstruction of left ventricle during a cardiac cycle was performed. The main work and achievements are as follows: 1 preprocessing of left ventricular ultrasound image. In this paper, several common nonlinear filtering methods, median filter and Lee filter, are synthesized. Morphological filtering is used to filter the left ventricular ultrasound image. The morphological filtering method is selected as the image preprocessing method through the comparative analysis of the results. 2 the registration of left ventricular MR images was studied. A novel image registration algorithm combining Radon transform and power spectrum is proposed for cardiac magnetic resonance imaging. The edge detection process of the algorithm not only filters out some noise interference, but also reduces the input data greatly. The feature of useful information and the rotation invariance of power spectrum are extracted quickly by using Radon transform to reduce the dimension. The registration of left ventricular MR image is well realized. 3 two methods of left ventricular ultrasound image segmentation were proposed. The first method is to accurately locate the left ventricular wall and extract the edge contour of the left ventricular wall in torsional motion by using the time-independent pulse coupled neural network. The second method uses the projection method and the C-V level set segmentation method to realize the automatic segmentation of the left ventricular ultrasound sequence image, which provides excellent materials for the further image registration. The algorithm firstly combines projection method to locate the position of left ventricle rough, then subdivides the location area by C-V level set method. 4 left ventricular motion tracking and three-dimensional reconstruction were realized. The edge of the segmented left ventricular ultrasound time series images was tracked with artificial markers and the motion parameters were measured and analyzed. The three-dimensional surface reconstruction of left ventricular cavity contour was carried out by using the shortest diagonal method under OpenGL platform. The three-dimensional reconstruction of left ventricle based on contour splicing is realized. The proposed heart motion tracking algorithm based on image processing is highly feasible. The image registration and image segmentation algorithms involved are original and have certain clinical application value. It is hoped that the segmentation and motion tracking algorithms will be further improved and applied to clinical analysis as soon as possible.
【学位授予单位】：重庆大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：R310;TP391.41

【参考文献】