Although it is a common problem, the diagnosis of sleep disorders, in general, is not easy : it involves sleeping in a laboratory connected to devices that track respiration, heart rate, movement and brain activity … Followed by expert data analysis. But now, a new technique that uses radio signals and algorithms could end the “sleep laboratory” forever , ushering in a revolution in these types of diagnostics.
It is a tool that adapts a convulsive neural network (CNN) to extract specific characteristics of the different stages of sleep, through radio frequency spectrograms. That is, the device can perform this “sleep reading” job located at a certain distance from the subject , which would avoid the annoying wiring currently required for monitoring. And without the need to go to a hospital.
How does it work?
The device, which must be installed in the bedroom , emits radio waves, similar to those of mobile phones and Wi-Fi router , measuring the return signal. The system is based on the previous monitoring of radio frequency sleep through the use of three machine learning algorithms to analyze respiration and pulse, and thus be able to identify the stage of sleep: light, deep, REM or wakefulness.
An algorithm uses a type of neural network common in image recognition to analyze the spectrograms, or snapshots, of the data. Another uses a type of neural network used to track temporal patterns to observe the dynamics of sleep stages . Finally, a third party refines the analysis to make it more generalizable across people and settings.
Researchers have enabled the tool for about 70,000 30-second sleep intervals and tested it on about 20,000. The measurement was compared to an EEG system that was as proficient as humans. The new system identified the stages of sleep with 80% accuracy , compared to 64% for the best radio frequency method available previously, the researchers reported. If the system hits the market, doctors may soon be able to easily diagnose sleep disorders.
The researchers believe that this work marks a milestone in sleep monitoring . The characteristics of this new predictive tool could also be applied to other health areas , in the diagnosis of other conditions in which the variations between subjects pose a great challenge for physicians.
Sleep disorders are a frequent problem in Western society, as much as that between 25 and 35% of the adult population suffers from transient insomnia , and up to 4% of men and 2% of women suffer from sleep apnea , according to data from the Spanish Society of Neurology (SEN). Data that would not be relevant were it not for the fact that little sleep and poor sleep quality are associated with multiple health problems , such as diabetes and, in general, shortens life expectancy.
