REDUNDANCY IN FUNCTIONAL BRAIN CONNECTIVITY FROM EEG RECORDINGS

The concept of redundancy is a critical resource of the brain enhancing the resilience to neural damages and dysfunctions. In the present work, we propose a graph-based methodology to investigate the connectivity redundancy in brain networks. By taking into account all the possible paths between pairs of nodes, we considered three complementary indexes, characterizing the network redundancy (i) at the global level, i.e. the scalar redundancy (ii) across different path lengths, i.e. the vectorial redundancy (iii) between node pairs, i.e. the matricial redundancy. We used this procedure to investigate the functional connectivity estimated from a dataset of high-density EEG signals in a group of healthy subjects during a no-task resting state. The statistical comparison with a benchmark dataset of random networks, having the same number of nodes and links of the EEG nets, revealed a significant (p < 0.05) difference for all the three indexes. In particular, the redundancy in the EEG networks, for each frequency band, appears radically higher than random graphs, thus revealing a natural tendency of the brain to present multiple parallel interactions between different specialized areas. Notably, the matricial redundancy showed a high (p < 0.05) redundancy between the scalp sensors over the parieto-occipital areas in the Alpha range of EEG oscillations (7.5–12.5 Hz), which is known to be the most responsive channel during resting state conditions.