A sleep monitor employing non-intrusive radar technology and machine learning can provide automatic sleep stage classification in healthy subjects. Polysomnography (PSG) is the medical gold standard for objective sleep measurements. The present study validates a non-contact sleep sensor using radar technology and machine learning against PSG. The results show that the monitor can provide automatic sleep stage classification with a precision close to PSG in a sample of healthy, mostly young subjects.
The present study demonstrates the ability of the novel sleep monitor to estimate sleep and wake periods in a healthy population. Compared to the manually scored PSG, it scored sleep/wake highly accurate, with 97 percent of sleep periods and 72 percent of wake periods scored correctly.
Looking at the sleep phases specifically, Somnofy accurately detected light sleep in 75 percent of time periods, deep sleep in 74 percent and REM in 78 percent of time periods compared to PSG. The overall average disagreement was 1 minute for light sleep, 12 minutes for deep sleep and 11 minutes for REM sleep compared to how much the manual PSG scores agreed.
> From: Toften, Sleep Med 75 (2020) 54-61 . All rights reserved to Elsevier B.V. Click here for the online summary.