A Graph Signal Processing Approach to Study High Density EEG Signals in Patients with Disorders of Consciousness

Graph signal processing (GSP) is a novel approach to analyse multi-dimensional neuroimaging data, constraining functional measures by structural characteristics in a single framework (i.e. graph signals). In this approach, functional time series are assigned to the vertices of the underlying weighted graph and GSP analysis is performed in each time point of the signal. Here we used GSP to study local brain connectivity changes in patients with disorders of consciousness based on resting state high density electroencephalography (hdEEG) recordings. Total variation of the graph signals is a measure of signal smoothness over the underlying graph. In this study, we constructed the underlying graph based on the geometrical distances between each electrode pairs in such a way that local smoothness of the signal can be studied. Total variation analysis in alpha-band showed that in the pathological states of altered consciousness, local short range communication of brain regions in this frequency band is stronger than in healthy states which shows that information is segregated in local regions in patients with disorders of consciousness.