Modularity analysis of brain network under real-time working memory feedback training

Abstract

Working memory (WM) is particularly important for higher cognitive tasks. Previous studies have shown that there are several brain networks under WM task or training, however, it is still unknown how many networks are involved in WM. In this paper, we utilize the method of modularity in the graph theory to explore the module distribution and the degree of coupling of the brain network under the real-time functional magnetic resonance (rtfMRI) WM training. The results suggest that there are four modules under the WM training, and the training changes the modularity of networks significantly. In addition, we further investigate the association between the WM capacity and the modularity and find no significant correlation between the variation of modularity and behavioral changes, which suggests that the modularity of brain network doesn't directly affect the WM performance. Such method of modularity provides a new perspective on network researches under WM.