Ge Bao, Guo Lei, Zhu Dajiang, Zhang Tuo, Hu Xintao, Han Junwei, Liu Tianming
Inf Process Med Imaging. 2013;23:692-704. doi: 10.1007/978-3-642-38868-2_58.
Modeling the human brain as a network has been widely considered as a powerful approach to investigating the brain's structural and functional systems. However, many previous approaches focused on a single scale of brain network and the multi-scale nature of brain networks has been rarely explored yet. This paper put forward a novel framework to construct multi-scale common networks of brains via multi-scale spectral clustering of fiber connections among DICCCOLs. Specifically, the recently developed and publicly released DICCCOLs provide the nodal structural and functional correspondence across individuals, and thus the employed multi-scale spectral clustering algorithm divided the DICCCOL landmarks and their connections into sub-networks with correspondences on multiple scales. Experimental results showed the promise of the constructed multi-scale networks in applications of structural and functional connectivity mapping. As an application example, these multi-scale networks are used to guide the identification of multi-scale common fiber bundles across individuals and to facilitate the bundle's functional role analysis, which could enable other tract-based and network-based analyses in the future.
将人类大脑建模为一个网络已被广泛认为是研究大脑结构和功能系统的一种强大方法。然而，许多先前的方法集中在大脑网络的单一尺度上，而大脑网络的多尺度性质尚未得到充分探索。本文提出了一种新颖的框架，通过对DICCCOLs之间纤维连接的多尺度谱聚类来构建大脑的多尺度公共网络。具体而言，最近开发并公开发布的DICCCOLs提供了个体间节点的结构和功能对应关系，因此所采用的多尺度谱聚类算法将DICCCOL地标及其连接划分为具有多尺度对应关系的子网。实验结果表明，所构建的多尺度网络在结构和功能连接映射应用中具有前景。作为一个应用示例，这些多尺度网络用于指导个体间多尺度公共纤维束的识别，并促进束的功能作用分析，这可能会在未来推动其他基于束和基于网络的分析。
