PhD Scientific Days 2023

Budapest, 22-23 June 2023

Mental Health Sciences I.

Parameters of the EEG spectrum as quantifiers of human sleep

Schneider Bence1
1 Institute of Behavioural Sciences, Semmelweis University

Text of the abstract

Introduction: While the field of quantitative sleep research is continuously expanding due to the wide availability of various electroencephalographic (EEG) methods, some of the most fundamental concepts and practices used today are still lacking quantitative description and definition. Such basic concepts include the stages of sleep, that are still identified visually by trained experts employing a number of predefined, consensual rules.

Aims: We try to find a set of mathematically defined and automatically computable indicators that can yield meaningful information about human sleep and which might provide an alternative to the rule-based, manual sleep-stage scoring, that is the standard today.

Methods: Sleep EEG contains two qualitatively different components: neural oscillations and a stochastic background activity, that manifest themselves in the power spectrum as spectral peaks and a decaying power-law respectively. We describe the EEG spectrum based on this physiologically-founded mathematical model and analyze the parameters. The power-law component can be described by the spectral slope and intercept, while the spectral peaks are characterized by peak frequency, power and bandwidth. After exploring general effects of sleep-stage, brain region and age, the temporal evolution of the parameters were extracted using a moving-window technique and their dynamics was analyzed.

Results: The above method was applied to an EEG-dataset containing the whole-night recordings of 251 subjects from a wide age range (4-69 yrs). The spectral slope showed significant and consistent effects with respect to sleep stages and reflected sleep depth. The continuous analysis found that the spectral slope exhibits a roughly 90-minute period, the known average sleep-cycle length of healthy humans. The slowing of the sleep cycle with aging and the unique persistence of sleep states in children was found as well. The method was also applied to a small data set of healthy and psycho-physiological insomnia patients where the dynamics of the spectral slope showed clear qualitative differences.

Conclusion: The spectral parameters are good candidates for the objective indexing of human sleep as they are sensitive to sleep dynamics reflecting several known phenomena. Furthermore, they might provide a basis for the quantitative description of the sleep process and an alternative to classical visual scoring.