Impact of gray matter signal regression in resting state and language task functional networks

A network analysis of the resting state (RS) and language task (LT) of fRMI data sets is presented. Specifically, the analysis compares the impact of the global signal regression of gray matter signal on the graph parameters and community structure derived of functional data. It was found that, without gray matter signal regression (GSR), the group comparison showed no significant changes of the global metrics between the two conditions studied. With gray matter signal regression, significant differences between the global (local) metrics for the conditions were obtained. The mean degree, the clustering coefficient of the network and the mean value of the local efficiency were metrics with significant changes. The community structure of group connectivity matrices was explored for both conditions (RS and LT) and for different preprocessing steps. When gray matter signal regression was performed, small changes of the community structure were observed. Approximately, the same regions were classified in the same communities before and after GSR. This means, that the community structure of the data is weakly affected by this preprocessing step. The modularity index presented significant changes between conditions (RS and LT) and between different preprocessing pipeline.