CIF:Small: A Signal Processing Approach to the Analysis of Time-Varying Functional Networks of the Brain

复杂网络理论已被证明是一个通用的框架来表示和分析包括社会科学、生物学和信息系统在内的许多学科中的关系数据。从这些发展中受益的一个特殊应用是认知神经科学。当代神经成像技术提供了越来越高的空间和时间分辨率的神经记录，产生了丰富的多通道数据集，可以用来详细描述大脑的功能连接模式。最近的研究证明，跨时空尺度的神经整合在广泛的认知和执行过程以及神经疾病和精神病理的表现中起着重要作用。目前对功能性脑网络的描述仅限于全球和静态测量，这些测量量化了受试者、大脑区域和时间的平均活动。本研究通过开发一个信号处理框架来研究基于多通道脑电图（EEG）数据的功能性脑网络的时变性质，从而更好地理解认知控制，从而解决了这一问题。本研究发展了三种主要的方法来研究大脑功能连接的动态性质。首先，采用时变的同步度量和统计假设检验来构建跨时间和频率的稀疏加权图。其次，基于信息论准则，提出了分层多主体聚类算法，用于识别时变社团结构；第三，开发了一个统计信号处理框架，以总结具有几个代表性网络的动态网络活动，并识别暂态和平稳活动。最后，本研究将应用于认知控制的研究，以确定控制认知和知觉的途径，并了解精神病理的基本网络原因。

Complex network theory has proved to be a versatile framework to represent and analyze relational data in many disciplines including the social sciences, biology and information systems. One particular application that has benefited from these developments is cognitive neuroscience. Contemporary neuroimaging techniques provide neural recordings with increasing spatial and temporal resolution yielding rich multichannel datasets that can be exploited for detailed description of functional connectivity patterns in the brain. Recent research provides evidence that neural integration across various spatial and temporal scales plays an important role in a wide range of cognitive and executive processes as well as in the manifestation of neural diseases and psychopathologies. The current characterizations of functional brain networks are limited to global and static measures that quantify average activity across subjects, brain regions and time. This research addresses this problem by developing a signal processing framework to study the time-varying nature of the functional brain networks based on multichannel electroencephalogram (EEG) data for a better understanding of cognitive control. This research develops three major approaches to study the dynamic nature of functional connectivity in the brain. First, time-varying measures of synchrony along with statistical hypothesis testing are implemented to construct sparse weighted graphs across time and frequency. Second, hierarchical multiple subject clustering algorithms are developed with information theoretic criteria to identify time-varying community structure. Third, a statistical signal processing framework is developed to summarize dynamic network activity with a few representative networks and to identify transient and stationary activity. Finally, this research is applied to the study of cognitive control for identifying the pathways that control cognition and perception, and in understanding the basic network causes of psychopathologies.