Unsupervised Characterization of REM Sleep Behavior Disorder from Home-Based EEG Recordings

Rapid-eye movement (REM) Sleep Behavior Disorder (RBD) is a parasomnia characterized by the loss of physiological muscle atonia during REM sleep, often resulting in dream enactment and violent behaviors. Idiopathic RBD is recognized as a prodromal marker of alpha-synucleinopathies, with conversion rates reaching 73–96% within 12–14 years. Despite its clinical relevance, diagnosis relies on polysomnography (PSG) in specialized sleep laboratories, limiting its accessibility and scalability, and its heterogeneity remains underexplored. To support remote health monitoring and early neurological assessment, this study proposes a sleep-stage-independent, unsupervised Machine Learning framework based on a single EEG channel collected during home-based PSG. In a dataset of 32 individuals with RBD, the method consistently identified two clusters with distinct neurophysiological profiles. These data-driven subgroups also showed differences in the REM Atonia Index (RAI), a quantitative marker of REM muscle atonia and RBD severity, supporting the interpretability of the identified neurophysiological patterns. These findings highlight the feasibility of low-complexity, unsupervised EEG analysis for characterizing RBD heterogeneity, suggesting the feasibility of scalable telehealth solutions for continuous monitoring and early detection of neurodegenerative risk.