Go with the Winner: Optimizing Detection of Modular Organization Differences in Dynamic Functional Brain Networks

摘要

The modular structure of human brain network(s) is well established. Despite numerous and increasing studies that examine brain's modular organization based on various measures of neural synchrony, it is not known yet how to qualify the employed descriptors in terms of the resulting functional community structure. A methodology is introduced here that facilitates the selection of best synchronization measure based on the comparison between two experimental conditions. Our method is presented using data from a multi-trial ERP paradigm (where the same task is performed in an attentive/passive mode) and in a time-varying exploration set up. The functional interactions are quantified at the level of EEG sensors through descriptors that differ regarding the nature of functional dependencies sought (linear vs. nonlinear) and regarding the specific form of the measures employed (amplitude/phase covariation). The resulting functional connectivity graphs (FCGs) are analyzed with an iterative clustering algorithm, and the emerging modular structures enter an appropriate time-varying discriminant function. Our results show that phase synchrony plays a crucial role in the segregation into distinct functional domains during the attentive condition in the frequency range that includes ? and a1 band (4 -- 10 Hz). Finally, by adopting Participation Index (PI), task-specific hub regions can be recognized from the optimally detected functional communities.