An EEG-based methodology for the estimation of functional brain connectivity networks: application to the analysis of newborn EEG seizure

This study presents a new methodology for obtaining functional brain networks (FBNs) using multichannel scalp EEG recordings. The developed methodology extracts pair-wise phase synchrony between EEG electrodes to obtain FBNs at δ, θ, and α -bands and investigates their network properties in presence of seizure to detect multiple facets of functional integration and segregation in brain networks. Statistical analysis of the frequency-specific graph measures during seizure and non-seizure intervals reveals their highly discriminative ability between the two EEG states. It is also verified by performing the receiver operating characteristic (ROC) analysis. The results suggest that, for the majority of subjects, the FBNs during seizure intervals exhibit higher modularity and lower global efficiency compared to the FBNs during non-seizure intervals; meaning that during seizure activities the networks become more segregated and less aggregated. Some differences in the results obtained for different subjects can be attributed to the subject-specific nature of seizure networks and the type of epileptic seizure the subject has experienced. The results demonstrate the capacity of the proposed framework for studying different abnormal patterns in multichannel EEG signals.