基于特征约束的三维光流模型的扩散张量图像配准研究

项目名称： 基于特征约束的三维光流模型的扩散张量图像配准研究

项目编号： No.61273261

项目类型： 面上项目

立项/批准年度： 2013

项目学科： 自动化技术、计算机技术

项目作者： 文颖

作者单位： 华东师范大学

项目金额： 81万元

中文摘要： 脑图像配准是脑图像分析的核心前提。目前，高维非线性空间的扩散张量图像（Diffusion Tensor Images,DTI）配准研究刚刚起步，主要研究思路为特征配准和灰度变换，且分析对象是正常的脑图像，模型的精确度和拓展性有限。考虑到脑成像的对象从新生儿到成人，成像质量千差万别，本课题将结构特征配准与光流理论相结合，提出了基于特征约束的高维光流算法的DTI配准思路。拟对DTI脑组织分割提取解剖结构特征并建立基于光流约束机制的特征描述配准模型，将该特征配准模型融入高维光流配准模型中，从而拓展传统光流模型，实现整体灰度配准和结构特征配准相结合的机制，然后对张量重定向研究，完成DTI配准。期望通过本课题的研究，能在光流法脑图像配准的理论上有较大的突破，不仅对正常脑图像而且对质量较低的新生儿图像和有缺损的脑结构图像配准上有所提高，为脑图像的高精度、快速及鲁棒性配准提供理论基础和关键技术。

中文关键词： 扩散张量图像；脑图像配准；高维光流；特征约束；

英文摘要： Brain image registration is the key to brain image analysis. At present, the high-dimensional nonlinear spatial registration of diffusion tensor image has just started. The idea of research is mainly based on feature match and gray transformation and the research object is often based on the normal brain image. In addition, the accuracy and expansibility of the model is limited. Given the various image quality and the image obtained from adult to baby, we propose a diffusion tensor image registration based on feature descriptor match combined with three dimensional optical flow algorithm. We obtain anatomy structures feature by brain tissue segmentation algorithm and design a feature descriptor match model based on constrants of optical flow. Then, we integrate feature match model into three dimensional optical flow algorithm to expend the traditional optical flow model, thus we achieve the mechanism of the global and structural match. We implement diffusion tensorm image registration based on the above model and tensor reorientation. Through this research, we anticipate it can make many breakthroughs on the theory of brain image registration, and it can registrate not only for normal brain image but also for baby's brain image and lesion brain image. And the reasearch also can provide theory base and key techni

英文关键词： Diffusion tensor imaging；Brain image registration；High dimensional optical flow；Feature constraints；