图像配准算法及其在功能磁共振图像预处理中的研究

发布时间：2018-04-08 19:34

本文选题：功能磁共振成像　切入点：图像配准　出处：《上海电力学院》2017年硕士论文

【摘要】：近年来随着医学影像技术的迅速发展,在磁共振成像技术基础上发展起来的功能磁共振成像(functional Magnetic Resonance Imaging,fMRI)技术,由于其能够无创伤性地对脑功能进行准确的定位,并且具有较好的可重复性和可行性,因此得到广泛的关注和研究。然而,在这一类研究中由于成像设备、个体差异及头动的影响,使得采集的fMRI图像不能直接进行处理分析,必须先进行一系列预处理操作来消除这些因素的影响。其中,图像配准是较为关键的一步,其结果的好坏直接关系到最终脑功能定位的准确性。此外,在对fMRI数据建模分析时,模型的回归量之间往往存在一定的共线性,导致得到不准确甚至错误的统计结果。因此,正确对fMRI数据进行处理和分析是实现脑功能准确定位的基础和前提。本文以fMRI图像为研究对象,对f MRI数据处理和分析过程中存在的问题进行了深入的探讨和研究。通过揭示脑认知活动的深层机制,对许多重大脑疾病的诊断、治疗以及相关病理学、药理学研究都具有重要意义。本文的主要研究工作如下:首先,介绍了图像配准算法的基本概念和框架,从几何变换、特征空间、相似性测度以及搜索策略四个组成部分对配准过程展开阐述,并以实际的fMRI数据为例,利用SPM软件包对fMRI数据的一般处理过程和统计分析过程作了详细的阐述;其次,考虑到fMRI数据采集过程中被试头动的影响,提出了一种基于互信息的功能磁共振图像配准方法,并应用主轴法和多分辨率策略来提高配准的速度和精度;最后,针对fMRI数据分析过程中一般线性模型回归量之间存在的共线性问题,提出了一种正交化方法。通过将相关的两个回归量正交分解为两个独立的量,以此来消除其中一个回归量对另一个回归量在结果变量中的影响,从而得到更加准确的结果。
[Abstract]:In recent years, with the rapid development of medical imaging technology, functional Magnetic Resonance imaging of MRI (functional magnetic resonance imaging) technology has been developed on the basis of magnetic resonance imaging technology.And has good repeatability and feasibility, so it has been widely concerned and studied.However, in this kind of research, because of the influence of imaging equipment, individual difference and head movement, the collected fMRI images can not be processed and analyzed directly, so a series of preprocessing operations must be carried out to eliminate the influence of these factors.Image registration is a key step, and the result is directly related to the accuracy of the final brain function location.In addition, when modeling and analyzing fMRI data, there is always a certain collinearity between the regression quantities of the model, which leads to inaccurate or even incorrect statistical results.Therefore, the correct processing and analysis of fMRI data is the basis and prerequisite for accurate localization of brain function.In this paper, fMRI image is taken as the research object, and the problems existing in the processing and analysis of f MRI data are deeply discussed and studied.By revealing the underlying mechanism of brain cognitive activity, it is of great significance for the diagnosis, treatment, related pathology and pharmacological research of many major brain diseases.The main work of this paper is as follows: firstly, the basic concept and framework of image registration algorithm are introduced. The registration process is described from four parts: geometric transformation, feature space, similarity measure and search strategy.Taking the actual fMRI data as an example, the general processing process and statistical analysis process of fMRI data are described in detail by using SPM software package. Secondly, considering the influence of the head movement in the process of fMRI data acquisition,In this paper, a mutual information based functional magnetic resonance image registration method is proposed, and the principal axis method and multi-resolution strategy are used to improve the speed and accuracy of registration.Aiming at the problem of collinearity between regression variables of general linear model in the process of fMRI data analysis, a orthogonalization method is proposed.By decomposing two correlated regression variables into two independent variables, the influence of one regression quantity on the other regression quantity in the result variable is eliminated, and more accurate results are obtained.
【学位授予单位】：上海电力学院
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：R445.2;TP391.41

【参考文献】