Scientists have developed a machine learning algorithm that can predict a person’s level of daytime sleepiness based on their brain activity recorded during periods of wakefulness. The study, published in the journal Scientific Reports, used electroencephalography (EEG) data from individuals who underwent overnight polysomnograms, which records a range of physiological data as the patient sleeps. Researchers selected 31 participants with excessive daytime sleepiness and 41 without sleepiness, and used EEG data collected during periods of wakefulness to train the algorithm. The algorithm was 88% accurate in predicting severe sleepiness, as measured on the Epworth Sleepiness Scale, using features such as power spectral density and phase-amplitude coupling. Confounding clinical variables were also analysed.
This study represents important progress towards developing objective biomarkers of sleepiness that could be used in patients with sleep disorders, but ultimately may also have value in individuals with insufficient sleep in the wider population, said Dr. Lara V. Marcuse, lead author of the study. While these findings offer the potential for the development of a non-invasive method for diagnosing and monitoring sleep disorders such as sleep apnea and narcolepsy, Dr. Marcuse stressed that further research is necessary to validate the results and verify their generalisability.
