Smart device can predict death

American scientists have built a smart tool that monitors people intact for possible cardiac arrest while they are asleep.

The system, using an artificial intelligence algorithm, is based on a “smart” portable speaker (such as Amazon Alexa and Google Home) or on a smartphone to detect breath sound and automatically call for emergency assistance if needed .
It detects – from a distance of up to six meters – 97% of respiratory events predisposing to cardiac arrest, according to a new American survey presented in Digital Health Magazine “npj Digital Medicine”.

Every year millions of people all over the world are dying, often alone in their homes, when their heart stops beating. Approximately half of them appear shortly before agonizing breathing, as they struggle to get oxygen.

“This is a kind of noise that makes its uniqueness a good biomarker of sound that can be used to pinpoint heart attack,” said Dr. James Sansane, an assistant professor of anesthesia at the University of Washington Medical School.

In these cases, early cardiopulmonary resuscitation may double or even triple survival levels. This, of course, presupposes that someone is present at the time when the incident occurs, says the RES-ICU.

Researchers collected 162 emergency calls from emergency medical services between 2009 and 2017 and exported two and a half seconds of sound at the beginning of each breathing incident to create a total of 236 recordings.

The group stored the recordings on different “smart devices” – Amazon Alexa, an iPhone 5s and a Samsung Galaxy S4 – and used various artificial intelligence (engineering) techniques to enhance all data in 7,316 “positive” recordings.

At the same time, a “negative” set of data was created, ie sounds that were not categorized as agonizing breathing. These recordings contained typical sounds made by people in their sleep, such as snoring or obstructive sleep apnea, as well as usual random sounds such as barks, car corners, air conditioning noise, etc., often heard in a home.

In this way, the system was “trained” to distinguish the dangerous breathing disorders from the most innocent and not automatically notify the ambulance without cause.

The study shows that the new system correctly detects respiratory anomalies by 97% when the device is positioned within six meters of the source of the sound. The new algorithm could work either as an app or as a smartphone or smartphone operation that passively passes and monitors breath while people are asleep.

Researchers intend to commercialize new technology through the newly established Sound Life Sciences, a scientist at the University of Washington.