Aperiodic brain activity tracks temporal fluctuations during sleep: an (i)EEG study

Description

Introduction:Electrophysiological data exhibit an aperiodic (1/f-like) characteristic, where power decreases with increasing frequencies. The magnitude of this decrease, quantified as the aperiodic exponent, has been linked to states of consciousness. In this study, we delve deeper into aperiodic activity across different sleep stages, extending methodological approaches to address limitations of prior investigations, including
(a) reliance on simplistic estimation models, and
(b) inadequate temporal resolution for aperiodic parameter estimation.
Materials and Methods:To overcome these constraints, we explore modifications in both the frequency range and model complexity employed for aperiodic activity estimation during sleep. Additionally, we employ time-resolved analyses to capture fluctuations within and between sleep stages. We analyzed an intracranial electroencephalography (iEEG) dataset (n=91) collected from sleeping human patients and high-density EEG data obtained from 17 healthy individuals during overnight sleep. Spectral parameters were computed utilizing the specparam toolbox (formerly 'fooof'), involving comparisons of various model forms and computation of time-resolved estimates.
Results:Our findings reveal that switching from the narrowband frequency range (30-45Hz), often used in sleep literature, to a broader frequency range enhances model performance. Furthermore, adopting a more intricate model formulation that integrates the knee frequency - the point at which the spectral exponent changes - reveals that sleep-stage-specific alterations in the spectral exponent of iEEG data are primarily driven by changes in the knee frequency. In the realm of EEG, we present temporally detailed estimations of aperiodic activity, showcasing time courses that mirror transitions between sleep stages.
Conclusions: Overall, our results demonstrate that by expanding the model complexity and temporal resolution of the applied methods, aperiodic activity is able to capture changes in sleep dynamics. Additionally, we propose updated guidelines for aperiodic activity estimation in neural data, facilitating a more precise delineation of electrophysiological signals.
Acknowledgements: This work is supported by the Austrian Academy of Sciences (OEAW) and the Austrian Science Funds (FWF).

Period	22 Oct 2023
Event title	World Sleep 2023
Event type	Conference
Location	Rio de Janeiro, BrazilShow on map
Degree of Recognition	International