Temporally Resolved Analyses of Aperiodic FeaturesTrack Neural Dynamics during Sleep

Introduction:
The aperiodic (1/f-like) component of electrophysiological data dictates that power systematically decreases with increasing frequency. The exponent of such aperiodic signal has been shown to differentiate sleep stages. Previous studies typically measured this exponent over narrow frequency ranges and averaged across sleep stages.In this study, we performed fine-graine analysis of brain activity patterns during sleep, expanding the analysis to include wider frequency ranges and alternative models, such as detecting 'knees' in the aperiodic component, which reflect bends in the power spectrum indicating changes in the exponent.
Method:
We analysed aperiodic brain activity from two souces: an intracranial electroencephalography (iEEG) dataset (n=91) from human patients, as well as a high-density EEG dataset collected from 17 healthy humans across a full night of sleep. Spectral features were estimated using the specparam toolbox (formerly ‘fooof’), comparing model-forms and computing time-resolved estimates.
Results:
Our results suggest a changes in the processing timescales of the underlying physiological activity across sleep stages, offering new insights into the understanding of changes previously attributed to cortical excitation-inhibition ratio. In addition, our method allows detailed examination of brain transitions between sleep stages, revealing rich temporal dynamics beyond the more typical 30-second analysis windows. Furthermore, our approach delineates transient responses to stimuli, showing how rapid transitions in aperiodic activity reflect the processing of external events, and how this varies across sleep stages.
Conclusion:
These findings enhance our understanding of brain dynamics during sleep and paves the way for future work on time-resolved quantitative analysis of aperiodic activity in electrophysiological signals.