基于静息态fMRI数据的人脑功能连接研究

发布时间：2018-03-28 01:08

本文选题：静息态功能磁共振成像　切入点：脑功能子网络　出处：《南京理工大学》2014年硕士论文

【摘要】：静息态功能磁共振成像(functional megnetic resonance imaging, fMRI)反映了大脑在静息状态下的活动情况,现已被广泛用于研究人脑功能。本文主要采用静息态fMRI数据进行了两个方面的研究。首先,提出了一种基于稀疏表示构建脑功能子网络的方法。该方法将小波变换一致性定义为感兴趣区域(ROI, Region Of Interest)之间功能交互模式的测度,然后通过稀疏表示提取每个ROI的特征,最后运用谱聚类的方法进行启发式聚类。本文对10个健康被测试对象的静息态fMRI数据进行了分析,构建得到的脑功能网络具有8个子网络。结果表明,这8个子网络分别对应着大脑的视觉、运动感知等主要功能系统,与当前的神经科学的研究相一致。同时本文还使用了两种传统方法来构建功能交互子网络进行比较,相比之下,本文方法构建得到的子网络效果更好。本文研究的第二个内容是,提出了一种基于稀疏表示推断健康人群与轻度认知障碍(Mild Cognitive Impairment, MCI)患者的特征功能连接的方法。首先也是通过小波变换一致性计算两个ROI之间功能连接的度量,然后对每个被试使用稀疏表示进行字典学习与稀疏编码,最后通过统计的方法分别对健康被试对象和MCI患者提取共性的功能连接。为了使算法更加通用,本文对两个不同的数据集共63个被试对象进行了研究。我们将得到的特征功能连接采用SVM进行分类,结果令人满意。最后,为进一步验证该方法的有效性与健壮性,本文生成了一系列人造数据并对其执行本文的方法,结果显示本文的方法是有效并且健壮的。
[Abstract]:Resting state functional magnetic resonance imaging (functional megnetic resonance imaging, fMRI) reflects the situation of the brain in the resting state activity, it has been widely used to study brain function. This paper mainly adopts the resting state fMRI was studied in two aspects. Firstly, propose a sparse representation based method to construct the brain functional network the method of wavelet transform. The consistency is defined as the region of interest (ROI Region, Of Interest) measure between functional interaction patterns, and then through the sparse representation of each ROI feature extraction, finally using the spectral clustering method of heuristic clustering. In this paper, 10 healthy test resting state fMRI data object is analyzed that building has 8 sub network of brain function network. The results show that the 8 sub network corresponding to the brain's visual perception, the main function of motion system, and the current It is consistent with the research of neuroscience. Meanwhile, two traditional methods are used to construct functional interaction sub networks for comparison. In contrast, the subnet constructed by this method has better effect.
The second content is proposed based on sparse representation that healthy people with mild cognitive impairment (Mild Cognitive, Impairment, MCI) method of feature function of patients. The first is connected by measuring function between two ROI connection calculation of wavelet transform consistency, then said the dictionary learning and sparse encoding for each is the use of sparse, finally through the statistical method of healthy subjects and patients with MCI extraction of common functions connected. In order to make the algorithm more general, based on two different data sets of a total of 63 subjects were studied. We will get the features connected by SVM classification, with satisfactory results finally, to further verify the effectiveness and robustness of the method, we create a series of synthetic data and the implementation of this method, results show that this party The law is effective and robust.

【学位授予单位】：南京理工大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：R445.2;TP391.41

【参考文献】