Alexa could spot your cardiac arrest—by listening to your breathing

If you’re unlucky enough to have a cardiac arrest, your chances are not good. Of the roughly 400,000 people who experience one every year in the US, less than 6% survive. It is the leading cause of natural deaths. However, the likelihood of survival is much higher if you get help quickly: immediate resuscitation can double or triple your chances.

A new tool that uses the microphone in your smart speaker or smartphone to detect warning signs, and then calls for help on your behalf, could help boost survival rates.

The system, developed by researchers at the University of Washington, uses machine learning to identify the telltale gasping sound (known as agonal breathing) that people make when they’re struggling for air. This is an early warning sign for more than half of all cardiac arrests.

The researchers trained the system using clips of agonal breathing captured from 911 calls made in King County, Washington. They used 729 calls yielding 82 hours’ worth of recordings in total. They then trained it on other sounds you might hear in someone’s room, like snoring or noises associated with sleep apnea, to weed out any false positives.

They used two different sets of recordings: sleep sounds collected by 35 volunteers, and those from 12 patients who were participating in a sleep study because they suffered from snoring and apnea. These latter recordings produced some sounds similar to agonal breathing, helping to refine the tool’s accuracy.

“When we tested it on our system, we found a 0.2% false positive rate in the volunteer group and a 0.1% rate in the sleep study,” says Justin Chan, who led the research.

The system managed to correctly identify agonal breathing in 97% of instances, from up to 20 feet away.

The tool, described in npj Digital Medicine today, is still at the proof-of-concept stage, so it’s many years away from being available to the public, although the researchers plan to commercialize it eventually. “There’s lots of work we’d have to do before we use this at scale,” Chan said.

He also suggested that if it were implemented in real life, it would make sense for the system to emit a 15- or 30-second warning to users that the emergency services are about to be called, so they have the chance to cancel in the event it’s a false alarm. It’s intended to be used in bedrooms, as this is where most cardiac arrests happen inside people’s homes.

There are still several challenges to overcome before the system can be launched, chiefly privacy issues, according to Peter Chai, an assistant professor of emergency medicine at Brigham and Women’s Hospital in Boston.

“There are questions around what you do with ambient noise of others in a room, or if you’re gathering information from a phone’s microphone, or what you do with inadvertent recording,” he says.