RBDAct: Home screening of REM sleep behaviour disorder based on wrist actigraphy in Parkinson's patients

Background: REM sleep behaviour disorder (RBD) is a disabling, often overlooked sleep disorder affecting up to 70% of patients with Parkinson's disease. Identifying and treating RBD is critical to prevent severe sleep-related injuries, both to patients and bedpartners. Current diagnosis relies on nocturnal video-polysomnography, which is an expensive and cumbersome exam requiring specific clinical expertise. Objectives: To design, optimise, and validate a novel home-screening tool, termed RBDAct, that automatically identifies RBD in Parkinson's patients based on wrist actigraphy only. Methods: Twenty-six Parkinson's patients underwent two-week home wrist actigraphy worn on their more affected arm, followed by two non-consecutive in-lab evaluations. Patients were classified as RBD versus non-RBD based on dream enactment history and video-polysomnography. We characterised patients' movement patterns during sleep using raw tri-axial accelerometer signals from wrist actigraphy. Machine learning classification algorithms were then trained to discriminate between patients with or without RBD using actigraphic features that described patients' movements. Classification performance was quantified with respect to clinical diagnosis, separately for in-lab and at-home recordings. Results: Classification performance from in-lab actigraphic data reached an accuracy of 92.9+/-8.16% (sensitivity 94.9+/-7.4%, specificity 92.7+/-13.8%). When tested on home recordings, accuracy rose to 100% over the two-week window. Features showed robustness across tests and conditions. Conclusions: RBDAct provides reliable predictions of RBD in Parkinson's patients based on home wrist actigraphy only. These results open new perspectives for faster, cheaper and more regular screening of sleep disorders, both for routine clinical practice and for clinical trials.