Abstract:Functional connectivity study using fMRI (fcMRI) is an important non-invasive means to explore the brain functional network. Lots of methods of fcMRI have been successfully used in resting-state. In this paper, we proposed a data-driven fcMRI method to explore the fcMRI of task-induced state. The core of the approach is trying to get the intrinsic BOLD patterns by means of clustering algorithm based on time course of every single voxel. In view of large numbers of fMRI data, we proposed a fusion approach based on multi-clustering to explore global brain task-induced network by comparing the performance of k-means, fuzzy cmeans and hierarchical clustering. On two real fMRI datasets, the approach can discover a global network which consistent with some prior knowledge of cognitive neuropsychological. These results demonstrate the proposed approach is an effective fcMRI method for task-induced state, and the approach can partly overcome the limitation, such as noise data source and seed selecting problem, which are inevitable in the study of fcMRI. © 2011 Binary Information Press.

A fusion approach based on multi-clustering to explore brain functional connectivity network