人脑结构网络中不同的脑区具有不同的连接模式，其中核心节点脑区促进了不同节点间高效的信息通信，是人脑结构网络的重要组成部分。可是，人脑结构网络中的核心节点是否可以被分为不同的类型？如果可以，那么不同类型核心节点的空间分布如何？他们的结构和功能特征以及与认知行为之间的联系又有着怎样的共性和差异？对于这些人脑连接组学领域的重要问题，人们知之甚少。本论文使用多模态神经影像数据集，通过重建大尺度下的人脑结构网络，定义了人脑结构网络的三类核心节点，揭示了它们在微结构、功能以及认知层面的特性。考虑到脑网络分析和统计的高度复杂性，作者作为核心人员开发了具有自主知识产权的基于图论的脑网络分析工具包GRETNA，为神经影像连接组学领域的研究提供了便利。该论文建立的神经影像脑连接组学计算方法及工具可被用于研究发育、老化和神经精神疾病的脑网络研究。 人脑结构网络不同类型核心节点的识别及其空间分布。使用公开的美国人脑连接组学项目数据库等三个多模态神经影像数据集，我们重建了大尺度下的个体人脑结构网络，然后基于八种常见的节点度量在人脑中定义了核心节点区域，并辅以综合的可重复性分析。我们发现人脑结构网络中有三类核心节点，将其命名为聚集型核心节点、分布型核心节点以及连接型核心节点。空间分布上，这些不同类型的核心节点都主要位于默认模式系统；更为重要地，聚集型核心节点还额外地分布于视觉系统和边缘系统，分布型核心节点还额外地分布于额顶系统，连接型核心节点还额外地分布于感觉运动系统和腹侧注意系统。此外，所有三类核心节点都表现出较高的跨扫描间空间可靠性，并且在个体识别中具有较高的预测率（分别为100%、100%和84.2%），能够扮演人脑结构“指纹”的角色。综上，我们发现了人脑结构网络中具有三类核心节点，为阐明人脑连接组的拓扑机理提供重要依据。 人脑结构网络不同类型核心节点的微结构和功能特征。在人脑结构网络中定义了三类核心节点后，我们考察了这些核心节点的结构和功能特征，包括微结构组织、布线成本、功能系统间整合、认知灵活性以及拓扑易受攻击性。这些不同类型的核心节点表现出差异化的特征，可能支持了他们在人脑中分化的角色：聚集型核心节点具有最好的微结构组织和最长的三维流线长度（即纤维长度），分布型核心节点保有最高的三维流线消耗（即布线成本）和最大的拓扑易受攻击性，而分布型核心节点呈现出最强的功能系统间整合能力和最大的认知灵活性。此外，这三类核心节点在所有特征方面都表现得比非核心节点更好。总而言之，我们强调了人脑结构网络中类型分化的核心节点，其具有差异化的结构和功能特征，为阐明人脑结构网络中核心节点的拓扑角色和功能作用提供了重要依据。 人脑结构网络不同类型核心节点与多种认知功能的关系。为了进一步阐明不同类型人脑结构网络核心节点的认知功能，我们使用美国人脑连接组学项目神经影像和认知数据库，通过典型相关分析研究了三类核心节点指数与12个认知维度的行为指标之间的关联性。三类核心节点广泛地参与到多数认知功能当中；其中，聚集型核心节点与分布型核心节点更偏重于参与空间定位、工作记忆、流体智力、阅读解码以及词汇理解等认知功能，所牵涉的脑区也遍布从视听觉皮层到联合皮层的各个功能系统，而分布型核心节点更多地参与到自我调节的冲动性、认知加工速度、情景记忆和词汇情景记忆等认知功能中，所牵涉脑区也相对集中在联合皮层和高级功能系统中。总的来说，我们的研究强调了三类核心节点参与人脑认知功能的共性与差异，初步探索了不同结构核心节点在人脑认知架构中的作用。 基于图论的脑网络计算分析工具包GRETNA。作为核心人员开发了具有自主知识产权的基于图论的脑网络计算分析工具包GRETNA，该工具包包含以下核心特征：（1）GRETNA是一个开源的、基于MATLAB的跨平台（Windows和UNIX-like）工具包，其具有用户友好的图形界面（GUI）；（2）GRETNA允许研究者以并行计算的方式计算脑网络全局和节点指标；（3）在脑网络构建和分析部分，GRETNA在几步核心的操作上提供灵活的选项设置，包括设置不同的节点定义方式、网络中正负连接选择、网络类型以及阈值化方式；（4）GRETNA允许研究者进行网络全局、节点和连边上网络指标的统计比较，并可以计算这些指标和给定临床或者行为变量的相关性；（5）最后，GRETNA集成了静息态功能磁共振数据预处理和功能连接矩阵计算功能。基于一个包含54名年轻成年人静息态磁共振数据集，我们使用GRETNA进行了脑功能网络分析，并论证了人脑功能网络具有高效的小世界、高同配性、层级化以及模块化的组织结构，并且具有高连边的核心节点；这些发现在不同的分析策略下依然可靠。GRETNA正在为神经影像脑连接组学领域的脑网络分析提供功能强大、分析快速并且灵活易用的解决方案。目前，GRETNA可在NITRC网站上公开免费下载，累积下载次数达5677次。该软件包的支持论文已入选ESI TOP 1%高被引论文。

Network hubs that facilitate efficient information communication are one of the most vital component of human brain structural networks. However, very little is known regarding whether structural hubs could be classified into different categories. If so, how are differentially categorized structural hubs spatially distributed in the human brain? How do their functional and structural characteristics vary? How their associations with cognitive behaviors are? In this thesis, using multi-modal neuroimaging dataset, we identified three categories of hubs in the large-scale human brain structural networks and further examined their microstructural, functional and cognitive characteristics. Given the huge complexity of brain network analysis and statistics, we also developed a GRaph thEoreTical Network Analysis (GRETNA) toolbox for imaging connectomics. Identification and Spatial Distributions of differentially categorized structural hubs. Using three multi-modal neuroimaging data sets, we construct individual structural brain networks and further identify hub regions based on eight widely used graph-nodal metrics, followed by comprehensive reproducibility analyses. We show three categories of structural hubs in the brain network, namely, aggregated, distributed and connector hubs. Spatially, these distinct categories of hubs are primarily located in the default-mode system; more importantly, aggregated hubs are additionally distributed in the visual and limbic systems, distributed hubs are additionally distributed in the frontoparietal system, and connector hubs are additionally distributed in the sensorimotor and ventral attention systems. Futhermore, all three categories of hubs display high across-session spatial reliability and act as structural fingerprints with high predictive rates (100%, 100% and 84.2%) for individual identification. Collectively, we highlight three categories of brain hubs, which shed light on topological mechanisms of the human connectome. Structural and functional characticitics of categorized structural hubs. After identifying three categories of structural hubs in human brain network, we investigate their various structural and functional characteristics, involving microstructural organization, wiring costs, functional modular integration, cognitive flexibility and topological vulnerability. These categorized hubs exhibit diverse characteristics to support their differentiated roles: aggregated hubs retain the best microstructural organization and the lengest streamline length (i.e. fiber length), distributed hubs present the highest streamline cost (i.e. wiring cost) and the heaviest topological vulnerability, and connector hubs show the strongest functional modular integration and the greatest cognitive flexibility. Furthermore, these characteristics are better in all three categories of hubs than non-hubs. In conclusion, we highlight the distinct structural and functional charactistics of three categories of differential hubs, which shed light on the topological roles and functional significance of structural hubs in human brain. Categorized structural hubs involved in multiple cognition. To illuminate the involvement of categorized structural hubs in cognition, we adopt canonical correlation analysis (CCA) to examine the association between three categories of hub indices and diverse behavior metrics of twelve cognitive functions. Three categories of hubs are widely involved in most of cognitive functions. Particularly, aggregated and distributed hubs tend to be involved in spatial orientation, working memory, fluid intelligence, language/reading decoding and language/vocabulary comprehension, and the associated brain regions are located at multiple fuctional systems, from visual cortex to association cortices; by contrast, connector hubs tend to be involved in self-regulation/impulsivity, processing speed, episodic memory and verbal episodi memory, and the associated brain regions are centered on association cortices and high-order functional systems. In short, we preliminarily explore the association between three categories of hubs and multiple cognition, which suggest the functional roles of differentiated hubs in human brain cognitive architecture. GRaph thEoreTical Network Analysis (GRETNA) toolbox. As a core member, we developed the GRETNA package, which contains several key features as follows: (i) an open-source, Matlabbased, cross-platform (Windows and UNIX OS) package with a graphical user interface (GUI); (ii) allowing topological analyses of global and local network properties with parallel computing ability, independent of imaging modality and species; (iii) providing flexible manipulations in several key steps during network construction and analysis, which include network node definition, network connectivity processing, network type selection and choice of thresholding procedure; (iv) allowing statistical comparisons of global, nodal and connectional network metrics and assessments of relationship between these network metrics and clinical or behavioral variables of interest; and (v) including functionality in image preprocessing and network construction based on resting-state functional MRI (R-fMRI) data. After applying the GRETNA to a publicly released R-fMRI dataset of 54 healthy young adults, we demonstrated that human brain functional networks exhibit efficient small-world, assortative, hierarchical and modular organizations and possess highly connected hubs and that these findings are robust against different analytical strategies. With these efforts, we anticipate that GRETNA will accelerate imaging connectomics in an easy, quick and flexible manner. GRETNA is freely available on the NITRC website with 5,677 total downloads.