Abstracts

Rationale: A better understanding of normal and aberrant cortical network function is crucial for developing more personalized diagnostics and treatments in epilepsy. Cortical network function in general, and information integration specifically, are believed to benefit from a network’s ability to integrate inputs over long timescales. Such long-range temporal correlations (LRTCs) generically arise when networks are poised at or near a critical point, i.e., near a phase transition between an active and an inactive dynamics regime. Monitoring LRTCs in cortical networks therefore provides a measure of information integration capabilities and, more generally, about the dynamical state of these networks [Meisel, 2020]. Previous research in non-human primates has indicated the existence of a gradient in LRTCs aligned with functional hierarchy [Murray et al., 2014], which may explain how inputs are integrated over increasingly longer timescales as a signal progresses higher along the hierarchical axis. However, whether LRTCs follow such a hierarchical gradient in the human brain also is still unclear. Further, whether LRTCs are impacted by different vigilance states, e.g., to provide a functional analog to the breakdown of information integration during slow wave sleep, is currently unknown. Lastly, the effect of antiepileptic drugs (AEDs), many of which impact network information processing, on these timescales remains to be determined. Here, we assessed LRTCs in human invasive EEG (iEEG) across different states of vigilance, as a function of AED load and across the hierarchical organization in cortical networks.

Methods: We assessed LRTCs in iEEG data from 17 patients with epilepsy undergoing presurgical monitoring (9 female, mean age = 27 ± 14 years, total duration 756.3 hours). LRTCs were determined at the individual channel level of power band fluctuation in the high gamma band (log power 56-96 Hz, 0.125 sec binning). LRTCs were calculated from the decay of power band autocorrelation functions (2 min segments, half-width at half maximum). Vigilance and slow wave sleep (SWS) states were classified as in Reed et al. 2017. Anatomical channel locations were determined from MNI coordinates. Evaluations included high and low AED drug days (24 h each).

Results: First, LRTCs were found to align along a functional hierarchy across different anatomical areas, thus electrophysiologically confirming similar findings from non-human primates in humans for the first time (Fig. 1). Second, LRTCs were found to break down during SWS compared to non-SWS (p< 0.02, two-sample t-test). Third, LRTCs decreased under AED administration (p< 0.02, two-sample t-test).