Abstracts

Rationale: Through the growing cross-disciplinary fields of engineering, computer science, and neuroscience, we are now encountering increasingly refined calculations and methods to determine cerebral functional connectivity. These methods span modalities from magnetic resonance imaging to magnetoencephalography to electroencephalography (EEG) with open source software dedicated to neuroscience, such as specialized MATLAB toolboxes. However, there is no current 'industry standard' for the optimal functional connectivity metric(s) in the literature in general, let alone the field of epilepsy. Our objective is to compare the relative validity of different EEG functional connectivity metrics against expected synchronization changes across different sleep-wake stages (i.e. awake, N1, N2, N3, REM sleep) using continuous EEG data derived from human patients in the epilepsy monitoring unit. Methods: Continuous interictal high density EEG recordings from 17 patients over 24-72 hours are manually sleep staged and subject to distinct algorithms for different EEG functional connectivity metrics in MATLAB to evaluate global synchronization. Chosen bivariate metrics include those in the non-directional linear time domain (e.g. correlation), non-directional linear frequency domain (e.g., coherence), non-directional linear time domain (mutual information), directional linear time domain (e.g., cross-correlation, Granger causality), directional linear frequency domain (e.g., phase slope index), directional non-linear time domain (e.g., transfer entropy). Chosen multivariate metrics include multivariate phase synchronization. These results are compared to expected global cerebral synchronization changes in different sleep-wake states.  Results: We expect to obtain drastically different results based on the choice of metric depending on its bivariate or multivariate nature, directionality or lack thereof, linear or non-linearity, and time or frequency domain. Expected global synchronization changes in different sleep-wake states are maximal synchrony in stage N3 sleep, maximal desynchrony in REM sleep, intermediate desynchrony in wakefulness, and intermediate synchrony in stages N1 and N2 sleep. The EEG functional connectivity metric(s) that best reflect these expected changes would be deemed the optimal metric(s).  Conclusions: Lack of an 'industry standard' for an EEG functional connectivity metric(s) poses formidable challenges to the valid interpretation of a rapidly increasing amount of scientific and medical literature on this topic in a wide variety of conditions, including epilepsy. Our work derived from human patients in the epilepsy monitoring unit will objectively, systematically, and rationally evaluate a collection of EEG functional connectivity metrics that will allow better understanding of the existing medical literature, and set the foundation for rigorous future studies into a promising and burgeoning aspect of epilepsy research stemming from the synergistic cross-pollination of the fields of engineering, computer science, and neuroscience.  Funding: No funding