Revealing a multiplex brain network through the analysis of recurrences

A multilayer approach has recently received particular attention in network neuroscience as a suitable model to describe brain dynamics by adjusting its activity in different frequency bands, time scales, modalities, or ages to different layers of a multiplex graph. In this paper, we demonstrate an approach to a frequency-based multilayer functional network constructed from nonstationary multivariate data by analyzing recurrences in application to electroencephalography. Using the recurrence-based index of synchronization, we construct intralayer (within-frequency) and interlayer (cross-frequency) graph edges to model the evolution of a whole-head functional connectivity network during a prolonged stimuli classification task. We demonstrate that the graph edges' weights increase during the experiment and negatively correlate with the response time. We also show that while high-frequency activity evolves toward synchronization of remote local areas, low-frequency connectivity tends to establish large-scale coupling between them.