Power-law parametrization of the sleep electroencephalography spectra: a new framework for modelling sleep-wake regulation
Horváth Csenge
Mental Health Sciences Division
Dr. Bódizs Róbert
SE Neurológiai és Pszichiátriai Klinika közös tanterme
2026-01-09 12:00:00
Magatartástudományok
Dr. Kovács József
Dr. Bódizs Róbert
Dr. Csukly Gábor
Dr. Tóth Attila
Dr. Juhász Gabriella
Dr. Rónai Katalin
Dr. Weiss Béla
Two process model is one of the most extensively used framework characterizing the regulation of sleep-wake states which lacks standardisable, non-redundant EEG markers. In my thesis, I aim to demonstrate, that by proper separation of the aperiodic and periodic part of the NREM sleep EEG Fourier spectrum, we obtain physiologically interpretable markers effectively reflecting known individual differences in sleep EEG and could serve as reliable indicators for homeostatic and circadian processes of sleep.
In Study 1, we demonstrated that spontaneous human brain activity measured by EEG during NREM sleep—in addition to rhythmic oscillations in specific frequency bands—, exhibits aperiodic, scale-free properties that follow a power-law scaling of the Fourier spectra. We derived four metrics using our parametrization procedures: spectral slope, the intercept, peak amplitude, and peak frequency in the spindle range. We showed that these measures, are sufficient to capture known age-, sex-, and IQ-related changes in the sleep EEG.
In Study 2, we analysed the first four NREM periods of sleep from 251 healthy human subjects (aged 4–69 years). We observed a flattening of spectral slopes, decrease in intercept, increase in spectral peak amplitude, and a U-shaped dynamic of peak frequencies in frontopolar regions. While the spectral slope reflected known age- and sex-related effects, the variability in its steepness was lower than that of SWA. Our findings suggest that combining scale-free and oscillatory measures of sleep EEG could offer composite indicators of sleep dynamics with minimal redundancy, potentially providing new insights into sleep regulation.
Finally, in Study 3, we validated the sleep homeostasis-related assumption regarding the spectral slope using a 35-hour sleep deprivation study. Spectral slope steepening effectively reflected changes in sleep depth due to sleep deprivation. Peak frequency during BS showed the expected overnight dynamics, with mid-sleep minima. BS timing of these minima significantly correlated with self-reported chronotype. However, sleep deprivation advanced the timing of the peak frequency minima in RS and reduced its correlation with chronotype.
Overall, our study highlights the spectral slope of sleep EEG as a marker of wake-sleep homeostasis and encourages further research on EEG-derived indicators of the circadian rhythm, particularly their interaction with the homeostatic process.
