Assessing the performance of a portable electroencephalographic sleep monitor against level 1 polysomnography

Study Objectives To assess the performance of a portable electroencephalography device for sleep monitoring against polysomnography. Method Fifty-six adults underwent one night of in-laboratory sleep recording with the Muse-S headband and simultaneous level 1 polysomnography. Muse-S data were scored by an automated sleep staging algorithm. A registered technologist, blind to the Muse-S automated sleep scoring, scored the polysomnography data. Results Good quality data were available for 47 (84 per cent) participants (53 per cent females; 20–71 years old; 17 per cent with sleep-related breathing disorder). Epoch-by-epoch analyses showed substantial agreement between the Muse-S and polysomnography (full night Cohen’s Kappa=0.76). Cohen’s Kappa were in the fair agreement range for non-rapid eye movement (NREM) 1, substantial agreement range for NREM2 and NREM3, and near-perfect agreement range for rapid eye movement sleep and wake. Accuracy ranged from 88 per cent to 96 per cent across all sleep stages, with a sensitivity of 79–92 per cent and a specificity of 90–99 per cent. Similar results were observed in the subgroup with sleep-related breathing disorder. On average, the Muse-S had higher mean values than polysomnography for total sleep time (+6 min), NREM3 (+15 min), rapid eye movement sleep (+6 min), and sleep efficiency (+1.5 per cent), and lower mean values for sleep latency (−3 min), wake after sleep onset (−3 min), and light sleep (−14 min). Conclusions When compared to standard polysomnography, the Muse-S performed well to measure sleep macroarchitecture. This portable device shows great potential as an accessible tool for sleep electroencephalography monitoring. More work is required to validate this tool in more diverse populations to ensure robustness across age, sex, neurological conditions, and sleep profiles.