SummaryThis project aims to investigate population-level sleep dynamics using NAKO SOMNOwatch recordings. Specifically, we will analyze objectively derived sleep–wake parameters to characterize sleep timing, continuity, and nocturnal organization in the general adult population. In addition to standard actigraphy-based measures, we will apply a previously developed convolutional neural network (CNN) to classify sleep stages (REM and non-REM) from the available recordings. Based on these sleep-stage time series, we will quantify transition probabilities between REM and non-REM sleep and estimate sleep cycle length and its variability across the night. These measures provide a compact statistical description of sleep architecture and its temporal organization that goes beyond conventional summary metrics. Another focus of the analyses will be the assessment of markers of sleep-disordered breathing using the SOMNOwatch respiratory airflow signal and ECG-derived dynamics. Where feasible, established or newly developed machine-learning models will be applied to derive indices of respiratory disturbances and to relate them to sleep-stage dynamics at the population level. Overall, the project extends established methodological work on sleep dynamics to a large population-based cohort, enabling a robust characterization of normative sleep-stage transitions and sleep-cycle structure in adulthood, as well as their association with markers of sleep-disordered breathing.

Keywords REM-NREM-transitions Somnowatch population-based-cohort sleep-cycles sleep-dynamics sleep-stages wrist-actigraphy

InstitutionsMartin-Luther-Universität Halle-Wittenberg, Universitätsmedizin Göttingen, Charité-Universitätsmedizin Berlin, Bar-Ilan University, Charite - Universitätsmedizin Berlin