SleepBCI: a platform for sleep quality assessment and memory enhancement based on automatic scoring

Abstract

Sleep is a complex physiological process responsible for several developmental, restorative and cognitive processes. Sleep related disorders diminish life quality, and their diagnosis relies on classifying 8-hour polysomnographic (PSG) recordings. Manual scoring is the gold standard for analyzing sleep and is a time consuming task that can only be performed by trained experts. We propose SleepBCI, a platform that integrates sleep quality analysis and closed-loop memory enhancement using automatic sleep scoring as its cornerstone. Firstly, real-time sleep staging is carried out through a convolutional deep neural network capable of achieving accuracies of 80-84% on a variable number of channels (1 to 6). Employing the neural networks’ output, we created an automated sleep quality report with objective metrics. This way, we characterized significant differences between healthy individuals and patients suffering from diverse sleep pathologies, such as breathing disorders, insomnia and narcolepsy. Furthermore, we implemented a brain-computer interface based on monitoring EEG activity in real time, intended to enhance previously learned memories during deep sleep. We validated it in a controlled study on healthy participants, who performed a word-sound association task before and after sleep. Recall performance after waking was significantly greater for reactivated cues compared with non-reactivated cues. The platform and results presented in this work serve as a proof-of-concept for future sleep technology. It lays the foundation for a take-home device capable of alleviating the workload of physicians, preventing at-risk populations (elderly or obese patients) from unnecessary PSG tests and finally, delivering therapy to slow cognitive impairment or dementia.