Deep Learning Enables Accurate Automatic Sleep Staging Based on Ambulatory Forehead EEG

We have previously developed an ambulatory electrode set (AES) for the measurement of electroencephalography (EEG), electrooculography (EOG), and electromyography (EMG). The AES has been proven to be suitable manual sleep staging self-application in in- home polysomnography (PSG). To further facilitate diagnostics various disorders, this study aimed utilize a deep learning-based automated approach EEG signals acquired with AES. present neural network architecture comprises combination convolutional recurrent networks shown achieve excellent scoring accuracy single standard channel (F4-M1). In study, model was re- trained tested 135 recorded recordings were conducted subjects suspected apnea or bruxism. performance learning evaluated 10-fold cross-validation using as reference. 79.7% ( $\kappa =0.729$ ) five stages (W, N1, N2, N3, R), 84.1% =0.773$ four light sleep, 89.1% =0.801$ three NREM, R). utilized able accurately determine based on channels measured is comparable inter-scorer agreement scorings between international centers. automatic AES-based could potentially improve availability PSG studies by facilitating arrangement self-administrated PSGs.