Towards fully automatic quantification of REM sleep without atonia according to the Sleep Innsbruck Barcelona (SINBAR) scoring method

Rapid eye movement (REM) sleep without atonia (RWA) is the polysomnographic hallmark of REM Sleep Behavior Disorder (RBD). The state-of-the-art methods to score RWA are visual-based. Recent international guidelines recommended the Sleep Innsbruck Barcelona (SINBAR) method for scoring RWA in 3-s mini-epochs. This method calls for scoring phasic EMG activity in the flexor digitorum superficialis (FDS) and “any” (i.e., tonic and/or phasic) EMG activity in the mentalis muscle. A semi-automatic algorithm scoring RWA according to this method is currently available in a commercial polysomnographic system (BrainRT, OSG, Belgium), however it still requires manual removal of EMG artifacts from expert scorers. This work proposes a novel method that, based on morphological aspects of EMG activity and machine learning (ML), discriminates activity from artifacts in the evaluation of RWA, thus allowing automatization for artifact correction