AI software helps emergency dispatchers spot cardiac arrests

By Jonah Comstock

A Danish AI system that can identify cardiac arrest from a 911 call 95 percent of the time might soon be coming to the US, according to a report in Fast Company

Corti, which launched from the company of the same name in 2016, is an example of how AI can augment, not supplant, human healthcare workers. The AI runs in the background of an emergency dispatch call, analyzing not only the caller's words but also background noises that might include the victim's breathing patterns. It culls that information for indicators that someone is suffering from cardiac arrest, then prompts the dispatcher to ask questions or walk the callers through activities like CPR that could stabilize the patient until the ambulance arrives.

When someone calls an emergency number (such as 911 in the United States or 112 in Denmark), the dispatcher has a difficult job. Once they have dispatched an ambulance, they must try to determine the nature of the injury — and do everything possible to stabilize the patient — by talking to family members or bystanders who are often panicked and generally lack medical training.

Corti uses machine learning to continually train itself to be better at picking up on signs of cardiac arrest, whether they're keywords in the caller's description or faint background sounds from the patient. According to a small study conducted by the company, trained dispatchers in Copenhagen alone could only identify cardiac arrest from phone calls 73 percent of the time. With the help of the algorithm, they knew 95 percent of the time. A second, larger study is forthcoming according to Fast Company.

Fast Company also reports that the company plans to make an announcement soon about bringing the technology to the United States. 

Check out the video below for a short demonstration of how the technology can help dispatchers.