Computational approaches to the study of REM Sleep Behavior Disorder

Parasomnias are a disorder that afflicts a part of the population, often can be easily solved, others instead can lead to more serious disorders, such as REM behavior disorder as it can be a warning premotor for neurogenerative diseases such as Parkison and Dementia with Lewy bodies. The study carried out makes it possible to analyze the various stages of sleep and subsequently extract the characteristics of each phase with EMG data from CMS (Sleep Medicine Center). The data, however, were noisy, so they were pre-processed with bandpass filters and notch filter because, due to the machines used, for each measurement it was noted there was a peak in 45 Hz, which was removed for all subjects. The distinction of the various stages of sleep was made by analyzing the hypnogram provided and, in particular, in the REM stage, RSWA (REM sleep without Atonia) scoring was calculated with the Montrèal and SINBAR method. The extracted and selected features of each stage were used for Machine Learning with the aim to automatically differentiate a unhealthy patient from a healthy one. Furthermore with the analysis of the phasic and tonic activity, the envelope of the tonic phase during REM sleep of the various subjects was studied, first by removing the phasic peaks and subsequently by enveloping the remaining tonic phase. It has been noted that in RBD subjects the envelope trend is modulated, but the reason is yet to be understood, but this can be a starting point for being able to compare the envelope trend of the other muscles to differentiate them or find common qualities.