Automatic classification of REM Sleep Behaviour Disorder exploiting Heart Rate Variability and Electromyography

Recent works highlighted that one of the fundamental roles of sleep is the removal of toxic metabolites produced during wakefulness. Consequently, a correlation was found between the quality of both rapid-eye movement (REM) sleep and deep sleep (i.e., slow wave sleep, SWS) and the development of neurodegenerative diseases. In this regard, the study of RBD has assumed great importance as it is considered a symptom that precedes, even by decades, the development of α-synucleinopathies, and a conversion rate to Parkinson's disease of around 90% has been found. The diagnosis of REM sleep behaviour disorder (RBD) is a lengthy and cumbersome process involving the analysis of polysomnography (approximately 8 hours of recording on more than 10 tracings) by a sleep expert. Consequently, the aim of this study is to develop a model capable of easing the physician's workload by performing an automatic binary classification between healthy and RBD patients. To date, some steps have already been taken in this direction often by exploiting electroencephalography (EEG) and electromyography (EMG) signals. However, following the indications of the latest research highlighting a relationship between heart rate variability (HRV) and RBD, this work aims to assess the predictive properties of the ECG signal, either singly or conjugated with the EMG signal, in detecting RBD. Different combinations of feature selection algorithms and machine learning models (preferred over deep learning due to the explainability of the results provided) were evaluated, obtaining accuracy values over 85%. This supports and confirms the research conducted on HRV and could further reduce the costs associated with the diagnosis of RBD, leading to the possibility of implementing screening procedures on the over-60 population. This would result in a very early diagnosis and, consequently, more effective treatment. Monitoraggio del disturbo comportamentale in sonno REM attraverso parametri fisiologici