Recognizing task-specific dynamic structure of the brain function from EEG

In this paper, we propose a novel framework aimed at identifying the spatio-temporal characteristics of the brain's functional networks specific to the task, from electroencephalography (EEG) recordings. Our analysis begins with segmenting the EEG data, from each trial, into intervals during which functional networks sustain their connectivity and spatial distribution in the cortex (the source space). The EEG data is then transformed into the source space for retrieving the spatial location of the cortical functional networks. Functional networks that are most recurrent across all segments and trials are then detected, and located by employing Boolean matrix factorization (BMF). Back-tracing the identified functional networks also reveals their dynamics that are specific to the task. Results for a visual odd-ball task are presented.