基于滑动窗口技术的脑网络动态特征矩阵构建方法研究

发布时间：2018-03-23 17:52

本文选题：滑动窗口技术　切入点：动态特征矩阵　出处：《昆明理工大学》2017年硕士论文

【摘要】：近年来,复杂网络理论、图论、计算机技术、生物科学技术及脑成像技术的快速发展,极大的推动了脑网络研究的进程,越来越多的科学工作者们致力于寻找和发现脑网络的结构、功能演变规律。通过对基于核磁共振技术的脑成像数据进行分析来构建脑网络并揭示大脑的运作机制,已经成为脑科学研究领域关注的重点。目前对脑网络的研究主要集中在静态特性研究,研究人员一般对一段时间内的脑成像数据信号进行均值化处理后再对脑区特性展开结构性、功能性或效应性的分析。实际上,大脑本身是一个随时在运行着的系统,基于核磁共振技术得到的大脑信号也是实时更新而且不断变化的,因此对人脑动态特性展开研究具有重要意义。本文针对人脑实时变化的特性,为了更好的观测和描述人脑网络的动态特征,提出了一种基于滑动窗口技术的动态特性分析方法来构建全脑动态特征矩阵,从而实现了对人脑动态特征的提取和描述。该方法利用滑动窗口技术的即时更新能力,首先把静息态功能磁共振成像数据采集区间上的血氧水平依赖信号由大时间窗口序列分解重构为每个采样点上的小时间窗口序列,从而构建连续时间点上的状态观测窗口;接着针对脑网络的分布性和活动并发性特征,将同步多维数据流的相关性分析算法应用于状态观测窗口内的数据分析,并构建得到了单状态观测矩阵;最后对全脑特征进行提取,构建得到了在整个数据采集区间上的全脑动态特征矩阵。该方法可以有效地将不同脑区的瞬时时间信号重组,并以特征矩阵的形式表现出来,从而得到全脑网络的动态特征。实验结果验证了该方法的可行性和适用性,为人脑网络动态特性的观测和描述提供了一定的理论基础。
[Abstract]:In recent years, the rapid development of complex network theory, graph theory, computer technology, biological science and technology and brain imaging technology has greatly promoted the progress of brain network research. More and more scientists are working to find and discover the structure and function of brain networks, by analyzing brain imaging data based on nuclear magnetic resonance (MRI) technology to construct brain networks and reveal the mechanism of brain function. Has become the focus of brain science research. At present, the research on brain network is mainly focused on static characteristics. Researchers generally analyze the structural, functional, or effectual characteristics of brain regions after averaging brain imaging data over a period of time. In fact, the brain itself is a system that is running all the time. The brain signals obtained by nuclear magnetic resonance (NMR) technology are also real-time updated and constantly changing, so it is of great significance to study the dynamic characteristics of human brain. In order to better observe and describe the dynamic characteristics of human brain network, a dynamic characteristic analysis method based on sliding window technology is proposed to construct the dynamic feature matrix of the whole brain. Thus, the extraction and description of the dynamic features of human brain are realized. This method utilizes the real-time updating ability of sliding window technology. Firstly, the blood oxygen level dependent signal in the interval of resting functional magnetic resonance imaging data acquisition is decomposed from the large time window sequence into the small time window sequence on each sampling point, and the state observation window on the continuous time point is constructed. Then, aiming at the distribution and concurrency of brain network, the correlation analysis algorithm of synchronous multidimensional data flow is applied to the data analysis in the state observation window, and the single state observation matrix is constructed. Finally, the whole brain feature is extracted, and the whole brain dynamic feature matrix is constructed in the whole data acquisition interval. This method can effectively recombine the instantaneous time signals of different brain regions and express them in the form of feature matrix. The experimental results verify the feasibility and applicability of the method and provide a theoretical basis for the observation and description of the dynamic characteristics of the human brain network.
【学位授予单位】：昆明理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：R318;O157.5

【参考文献】