Rapid Eye Movement Sleep Behavior Disorder Detection Using Smart Wristbands: A Preliminary Study

In recent years wearable commercial devices, such as smart wrist-bands, have become popular principally for sport activity monitoring. Among them, almost all the major marketed models are also able to evaluate the quality of a person's sleep, identifying the sleep stages and their durations. From a medical point of view, sleep stage patterns constitute an important preliminary exam in the diagnosis of sleep disorders, or possibly in the prevention of the same. Rapid Eye Movement (REM) sleep behavior disorder is one of the well-known sleep disorders, with a prevalence of around 2% in older adults, being a prodromal syndrome of neurodegeneration, and causing them Parkinson disorder (33–50%), multiple system atrophy (80–95%), and dementia with Lewy bodies (80%). It is advantageous therefore to have an automatic tool for REM sleep disorder detection that does not require the use of high-cost medical instrumentation (such as Polysomnography) or dedicated medical staff to analyze data. This paper proposes a preliminary study for the design and implementation of an algorithmic framework, based on Machine Learning approaches, able to detect and classify this pathology only analyzing the sleep stages retrieved by a low-cost and commercial wearable device. A benefit of the proposed system lies in the abstraction of the algorithmic pipeline with respect to the device used for sleep stage assessment. Preliminary results obtained on benchmark literature dataset demonstrate the effectiveness of the proposed pipeline at varying of different Machine Learning classifiers, achieving higher accuracy with Support Vector Machine.