This Software Uses AI to Identify Cardiac Arrests Faster Than Humans

Software that uses Artificial Intelligence (AI) to help emergency dispatchers identify cases of cardiac arrest will be tested throughout Europe this summer. The Danish startup, Corti, has developed the software that they say can help recognize out-of-hospital (ie in the home or in public) instances of cardiac arrest faster and more accurately than a human. 

The software has been tested in Copenhagen and this summer the pilot program will be rolled out in four other European cities as part of a partnership with the European Emergency Number Association (EENA). 

Software has the ability to save lives

The quick detection of cardiac arrest is vital in saving patients' lives. Every minute that passes during an arrest without treatment reduces the patient's chance of survival by 7-10 percent. Corti’s system listens into emergency calls and uses algorithms to identify both “verbal and non-verbal patterns of communication.” 

The software looks for cues such as the tone of the caller's voice and whether or not they are breathing. The system prompts the dispatcher to ask certain questions and guides them to a decision about whether to send an ambulance or instruct someone present to begin CPR. 

Algorithms look for patterns in calls that point to cardiac arrest

 It is important to note that cardiac arrest is not the same as a heart attack. Cardiac arrest occurs when an electrical fault causes the heart to stop beating. Despite sometimes being used interchangeably a heart attack happens when a blocked artery limits the circulation of blood around the body. 

Corti has amassed some very impressive stats in its testing period so far. Corti says that when tested on a database of 161,650 historical emergency calls, the software was able to accurately identify 93.1 percent of out-of-hospital cardiac arrests (or OHCAs) compared to 72.9 percent recognized by human dispatchers. 

Software faster and more accurate than human dispatch operators

Not only was the software more accurate it was faster, being able to identify key signs of cardiac arrest in 48 seconds compared to the 79 seconds it took for a human to do the same thing. Skeptics of the software's ability are waiting for the full report of this testing phase to be released to look at other numbers like the software's false positive rate, i.e. the number of times it incorrectly identifies cardiac arrests. 

Other people question the software's ability to make good decisions when faced with unknown situations or when faced with human nuances. For example, when someone is calling on behalf of a loved one, they often appear more confident that patient is breathing because they want to believe it's true. 

This accidental misinformation may cause the software to misunderstand the situation. But Corti’s developers say that the software never makes autonomous decisions and only acts as a guide for professionally trained humans to make the final decision.