乳腺动态增强MRI配准方法的研究

发布时间：2018-11-10 07:15

【摘要】：动态增强MRI由于具有无损伤性且可显示功能信息，在乳腺癌早期诊断和跟踪治疗方面得到了广泛应用。在DCE-MRI检测时，需要获取注射造影剂前后的患者乳腺影像。造影剂的使用能够改变肿瘤区的原始灰度，进而使得肿瘤与正常组织区分开来。一般而言，造影剂是通过静脉来注射的，这样随着跟踪诊疗过程中影像采集次数的增加，，造影剂的副作用会在患者体内不断累积。针对造影剂静脉注射方法的这一缺陷，从乳头注射造影剂的新型乳腺动态增强MRI方法应运而生。该方法虽然降低了造影剂的注射量，但由于在造影剂吸收初期，乳头附近的灰度变化剧烈，为图像的配准造成了极大困难。本文基于面网络配准技术对该特殊应用背景下的图像配准问题进行了探索，设计了以球谐分析为核心的面网络配准技术，并辅以适当的图像预处理和面配准后的体插值实现图像层面的配准。针对造影后图像靠近乳头的组织区域灰度接近背景而难以进行图像分割的问题，本文设计了基于灰度数学形态学的图像边缘修复方法，为较为精确的三维表面重建提供了保证。面网络的配准是由网络方位对准和节点间对应关系的确立这两步来实现的。本文首先利用保形球映射、规则采样和基于贪婪搜索的球谐配准实现造影前后的乳房面网络在方位上的配准，而后基于球谐分析的机理运用同球谐基底异球谐系数重构出具有节点间对应关系的造影前后的乳房面网络，从而获得了表面网络的位移场。针对图像层面的体配准，本文首先利用均值坐标插值技术在造影后面网络的内部插值出相应的体位移场，然后据此应用基于线性插值的图像复原技术实现了对原始图像的几乎所有像素点的形变位移操作，从而获得了组织灰度信息清晰可辨的配准图像。该图像具有造影前乳腺MRI的形态，且具有造影后乳腺MRI的组织灰度信息。该方法的提出为绘制动态增强（DCE）曲线，从而得到更为全面的诊断信息提供了坚实的基础。
[Abstract]:Dynamic enhanced MRI has been widely used in early diagnosis and treatment of breast cancer due to its noninvasive and functional information. During DCE-MRI testing, mammograms of patients before and after injection of contrast media are required. The use of contrast agents can change the original grayscale of the tumor area, which distinguishes the tumor from the normal tissue. Generally speaking, the contrast agent is injected through the vein, so as the number of image acquisition increases during the follow-up, the side effects of the contrast agent will accumulate in the patient's body. In view of the defect of intravenous injection of contrast agent, a new dynamic enhanced mammary gland MRI method for nipple injection was proposed. Although this method reduces the amount of contrast agent injected, it is very difficult for image registration because of the sharp change of gray level near the nipple in the early stage of absorption of contrast medium. In this paper, based on the surface network registration technology, the image registration problem under the special application background is explored, and the surface network registration technology with spherical harmonic analysis as the core is designed. The image level registration is realized with proper image preprocessing and volume interpolation after surface registration. In view of the difficulty of image segmentation due to the grayscale of the tissue area near the nipple after angiography, an image edge restoration method based on gray-scale mathematical morphology is designed in this paper, which provides a guarantee for the more accurate 3D surface reconstruction. The registration of surface network is realized by two steps: azimuth alignment of network and establishment of corresponding relation between nodes. In this paper, we first use conformal ball mapping, regular sampling and spherical harmonic registration based on greedy search to realize the azimuth registration of breast face network before and after angiography. Then, based on the mechanism of spherical harmonic analysis, the mammary face network with corresponding relationship between nodes is reconstructed by using the isobaric coefficient of the same spherical harmonic substrate, and the displacement field of the surface network is obtained. For the volume registration of image level, this paper firstly uses the mean coordinate interpolation technique to interpolate the corresponding position field in the interior of the network behind the angiography. Based on this, the image restoration technique based on linear interpolation is applied to realize the deformation displacement operation of almost all pixels of the original image, and the clearly identifiable registration image of the gray level information of the organization is obtained. The image has the shape of mammary MRI before contrast and grayscale information of MRI. This method provides a solid foundation for drawing dynamic enhanced (DCE) curves and obtaining more comprehensive diagnostic information.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：R737.9;R445.2

【参考文献】