Researchers at the University of Washington have used machine learning to teach an AI system to identify when someone is having a cardiac arrest. The system learned to identify agonal breathing, which occurs when someone gasps for breath during cardiac arrest, with a high degree of accuracy. The technology can be embedded into a variety of listening devices, such as smart speakers or smartphones, to alert authorities and loved ones to someone having a heart attack while they sleep.
Approximately half a million Americans die from cardiac arrest annually. Cardiac arrests often happen while someone is at home in bed. This is particularly dangerous, as there is likely to be no-one around, or no-one awake, to help.
Now, researchers have developed an AI system that can work through smart speakers or a smartphone to monitor for signs of a cardiac arrest while someone sleeps. The system listens for something called agonal breathing, which occurs in about 50% of people who experience a cardiac arrest, and patients who demonstrate this characteristic gasping often have a better chance of surviving.
“This kind of breathing happens when a patient experiences really low oxygen levels,” said Dr. Jacob Sunshine, a researcher involved in the study. “It’s sort of a guttural gasping noise, and its uniqueness makes it a good audio biomarker to use to identify if someone is experiencing a cardiac arrest.”
The researchers trained their system using recordings of agonal breathing from real 911 calls to Seattle’s emergency services. “We played these examples at different distances to simulate what it would sound like if the patient was at different places in the bedroom,” said Justin Chan, another researcher involved in the study. “We also added different interfering sounds such as sounds of cats and dogs, cars honking, air conditioning, things that you might normally hear in a home.” Recordings of normal sounds, such as snoring, served as a negative dataset to reduce the chance of the system misidentifying normal sleep as a cardiac arrest.
The machine learning approach paid off, as the system achieved an impressive 97% accuracy rate in identifying agonal breathing using a smart speaker or similar device placed as far away as 6 meters (20 feet) from the source of the sound.
“A lot of people have smart speakers in their homes, and these devices have amazing capabilities that we can take advantage of,” said Shyam Gollakota, a third researcher involved in the study. “We envision a contactless system that works by continuously and passively monitoring the bedroom for an agonal breathing event, and alerts anyone nearby to come provide CPR. And then if there’s no response, the device can automatically call 911.”
Study in journal npj Digital Medicine: Contactless cardiac arrest detection using smart devices