Department of Neurology, Nanfang Hospital, Southern Medical University, #1838 Northern Guangzhou Avenue, Guangzhou, 510515 Guangdong, China.
Neuroinformatics. 2009 Dec;7(4):233-44. doi: 10.1007/s12021-009-9056-z.
Phase synchrony has been proposed as a possible communication mechanism between cerebral regions. The participation index method (PIM) may be used to investigate integrating structures within an oscillatory network, based on the eigenvalue decomposition of matrix of bivariate synchronization indices. However, eigenvector orthogonality between clusters may result in categorization difficulties for hub oscillators and pseudoclustering phenomenon. Here, we propose a method of fuzzy synchronization clustering analysis (FSCA) to avoid the constraint of orthogonality by combining the fuzzy c-means algorithm with the phase-locking value. Following mathematical derivation, we cross-validated the FSCA and the PIM using the same multichannel phase time series of event-related EEG from a subject performing a working memory task. Both clustering methods produced consistent findings for the qualitatively salient configuration of the original network-illustrated here by a visualization technique. In contrast to PIM, use of common virtual oscillatory centroids enabled the FSCA to reveal multiple dynamical neural assemblies as well as the unitary phase information within each assembly.
相位同步被认为是大脑区域之间可能的通信机制。参与指数方法（PIM）可以基于双变量同步指数矩阵的特征值分解，来研究振荡网络中的整合结构。然而，簇之间的特征向量正交性可能导致主振荡器的分类困难和伪聚类现象。在这里，我们提出了一种模糊同步聚类分析（FSCA）的方法，通过将模糊 c 均值算法与锁相值相结合，避免了正交性的限制。在数学推导之后，我们使用来自执行工作记忆任务的受试者的相同多通道相位时间序列对 FSCA 和 PIM 进行了交叉验证。这两种聚类方法对于原始网络的定性显著结构都产生了一致的发现，这里通过可视化技术进行了说明。与 PIM 不同的是，使用通用的虚拟振荡质心使得 FSCA 能够揭示多个动态神经组合以及每个组合中的单元相位信息。
