Estimation of functional newborn's brain networks using EEG source connectivity

Background and Aim: Brain connectivity analysis has been well developed using different neuroimaging techniques and used to study normal brain functions as well as neurological disorders like epilepsy, autism, and schizophrenia. Among different neuroimaging techniques, EEG is more widely used due to its high temporal resolution, its non-invasiveness nature and relative easiness of use. The main drawback of EEG-based techniques is the effect of the head volume conduction which may result in false connections. In this study, a new methodology for studying brain functional connectivity networks (FCNS) is presented and applied to multichannel newborn EEG signals to reveal cortical networks during seizure and non-seizure EEG patterns. Methods The proposed methodology for the estimation of FCNs using multichannel EEGs consists of ۴ main steps. (۱) Pre-processing: the stationary wavelet transform is used to enhance the input multichannel EEG and decompose it into ۴ standard EEG frequency bands. (۲) Source localization: the minimum norm estimates algorithm is deployed to estimate the time-series at the source level corresponding to the input multichannel scalp EEG. (۳) Functional connectivity analysis: the multivariate phase synchrony measure based on the circular omega complexity is applied to the source time-series to find connectivity matrices. (۴) Visualization of brain networks using graphs: graphs are used to display a set of connections between brain regions by assigning nodes that represent places of electrodes and branches representing connections between the nodes. Results To (۱) present a new EEG-based technique for the estimation of FCNs, (۲) use it to visualize newborn's brain FCNs during seizure and non-seizure EEG patterns, and (۳) validate it using real multichannel newborn EEGs. Conclusion: The methodology was applied to a database composed of multichannel EEGs collected at the Helsinki University Hospital from infants admitted to the NICU. The experimental results reveal a statistical difference between the FCNs representing seizure and non-seizure states. The resulted graphs suggest that, in non-seizure states, the majority of connections tend to be between neighboring sources. In seizure states, on the hand, connections reach distant regions. Some differences were observed in graphs visualizing FCNs during seizure sates for different subjects. The methodology proposed in this study can therefore be used to identify neural networks involved in normal and pathological brain functions and can consequently aid clinicians in the diagnosis and estimation of neurodevelopmental disorders.

Keywords:

Functional brain networks , EEG signals , EEG source localization , brain connectivity analysis