The spectral slope of the EEG as an indicator of sleep depth and architecture
Schneider Bence
Mental Health Sciences Division
Dr. Bódizs Róbert
SE Magatartástudományi Intézet
2026-06-18 10:00:00
Magatartástudományok
Dr. Kovács József
Dr. Bódizs Róbert
Dr. Baradits Máté
Dr. Bálint Zoltán
Dr. Kovács Tibor
Dr. Tóth Brigitta
Dr. Eke András
Besides neural oscillations manifested as spectral peaks in the power-spectral den- sity (PSD) of the electroencephalogram (EEG), the brain constantly generates an aperiodic, fractal-like background activity that can be described by a power-law in the frequency domain. Several methods emerged recently that aim to separate the oscillatory and fractal components of electrophysiological signals and to give a parametric description at the same time.
In Study 1, we assessed how spectral parameters depended on sleep stages, age and brain region in a healthy population. Average PSDs were calculated for each sleep stage and EEG electrode per subject. Using the FOOOF method, the aperiodic and oscillatory spectral components were fitted and parameters of spectral slope and intercept, peak center frequency, peak power and peak bandwidth were extracted.
The spectral slope showed the most consistent and specific sensitivity to sleep stages, being the highest in wake and decreasing with the deepening of sleep, being minimal in slow-wave sleep, then increasing again in REM sleep. Significant effects and interactions with age and brain region were also revealed. Oscillatory parame- ters also showed sleep stage dependence, peak central frequencies being in the alpha band in wake, spindle range in NREM2 and SWS as expected, but some novel sim- ilarities were revealed between REM and NREM1 sleep, being dominated by beta activity after the removal of the aperiodic background. Furthermore, spindle peak power showed a non-linear effect, being lower in children and middle-aged adults, and most prominent in young adults.
In Study 2 the overnight dynamics of the slope reflected sleep structure, based on which a new definition of sleep cycles had been proposed. After smoothing and normalizing the slope time-series, local maxima were considered as markers of cycle turning points. The newly defined fractal cycles were compared to classical, man- ually annotated ones on datasets of healthy adults, children and major depressive disorder (MDD) patients, in most cases the cycles matching both in timing and dura- tion. Fractal cycle durations showed significant effects of age, increasing in children and decreasing in healthy adults, even after controlling the possible confounding effects of wake after sleep onset and REM latency. Fractal cycle durations differed in medicated MDD patients from their own unmedicated state and healthy controls. The effect of medication was significant only for REM-suppressing antidepressants.
Overall, the parametric description of EEG spectra showed potential to serve as a basis for the objective characterization of sleep states, paving the way toward a data- driven, biologically plausible evaluation of human sleep, providing an alternative to rule-based, manual scoring, furthermore, suggesting clinical applicability.
