These sleep monitoring innovations could make smartwatches better at tracking our slumber

2 minutes ago

Smartwatches may become better at tracking sleep. (Brian Ach/AP Images for Fitbit)

Scientists at Emory University and the Georgia Institute of Technology have created multiple new methods and systems for sleep monitoring, overcoming some limitations of previous approaches for tracking sleep disruptions with smartwatches or other wearables.

"I've been working on sleep for about 20 years now," said Gari Clifford, a professor of biomedical informatics and biomedical engineering at Emory University and the Georgia Institute of Technology, chair of Emory's Department of Biomedical Informatics, and lead author of the patent application, published by the World Intellectual Property Organization on March 11. "People who have disrupted sleep, or who get too little sleep, will go on to develop all sorts of complications that can be anything from cardiovascular to psychiatric. The big problem is measuring sleep."

The team's inventions stemmed from ongoing research on people with post-traumatic stress disorder, many of whom suffer from serious insomnia or other sleep disruptions. As part of the large, multi-institutional AURORA Project, Clifford and other scientists have pored over thousands of records from emergency departments across the country. They want to predict which trauma victims are most likely to develop PTSD. "The target is to have a total of 5,000 people that we've tracked for up to a year," Clifford explained. "So that could be as many as 100 billion heartbeats that we're processing from this cohort."

A hundred billion heartbeats, plus billions of other data points, is no small amount. "It starts to become quite cumbersome for a device on a wrist," Clifford said. "You're likely to miss a lot of data if you have to run these algorithms constantly."

These inventors set out to process such sleep-related information better: Their patent application includes a new, low-memory strategy for tracking sleep that relies solely on "change points," or the event timestamps for transitions from one sleep stage to another. "If you've got an algorithm running that just takes change points, and only saves these change points and the magnitude of the change, and then inputs that into an algorithm, you can compress that information down significantly," Clifford explained.

According to Clifford, the scientists' innovations include their technique for detecting those change points as well as the ways they combine separate channels and store data. "The idea of doing this on a wearable is novel because the entire system before it is novel," he added. "It doesn't matter if you do it on a wearable or an off-body system. We could take this same algorithm and apply it to off-body sensor systems, like cameras."

The inventions come as the sleep tech market has taken off, reaching a size of more than $10.9 billion in 2019, according to research firm Global Market Insights. And in recent years, clunky, expensive polysomnograms have increasingly given way to wearables such as Fitbits and Apple Watches. 

Yet according to Clifford, those wearables still have shortcomings: For one, they tend to misclassify wakeful periods as sleep. "If you've reclined on the sofa and watched TV late in the evening, you might find that your wearable on your wrist tells you that you were asleep, even though you clearly remember being awake," Clifford said. His innovations do not overestimate sleep, avoiding this issue.

Wearables also significantly underperform for individuals with sleep disorders. "The dirty secret is most of them work well when you have normal sleep structure, but when you have poor sleep structure, they don't actually work that well," Clifford said. Traditional algorithms were designed around populations with normal sleep structure, whereas his innovations were optimized on many different populations with poor sleep. This means that, among other things, they do not assume that people moving during sleep are awake. "Someone with restless leg syndrome moves a lot while they're asleep," he noted.

Clifford thinks his contributions could help provide a cheaper, less invasive alternative to sleep studies, which doctors commonly use to diagnose sleep apnea or other sleep disorders. 

"What would be better is if somebody could just give you a little patch," he said. "You go home, and you wear that for a few days in your home environment — your natural environment — and at the end of it, you put that patch in a box, or you synchronize it to your WiFi, and the data goes up to a central server, and you get a diagnosis back that says, 'This is the severity of your sleep apnea, and these are treatment choices.' And you can go back [to your doctor] and have that conversation."

While Clifford has not yet commercialized his patent application, he says he's interested in doing so in the future.

The application for this patent, "Systems and Methods for Detecting Sleep Activity," was filed on Sept. 4, 2020 to the World Intellectual Property Organization. It was published March 11, 2021 with the application number PCT/US2020/049392. The earliest priority date was Sept. 5, 2019. The inventors of the pending patent are Gari Clifford and Ayse Cakmak, Emory University; and Christopher Rozell and Adam Willats, Georgia Institute of Technology. The applicants are the Emory University Office of Technology Transfer and the Georgia Tech Research Corporation.

Parola Analytics provided technical research for this story.

We use cookies to improve your experience on our site and to show you relevant advertising.