How AI Can Detect Heart Attacks From Emergency Calls

I’ve written previously about how technology can monitor our voice for diseases, especially neurological ones such as Parkinson’s and Alzheimer’s.  The ultimate aim of such services is to provide a live diagnoses whilst we’re engaged in a regular conversation with our clinician.

A Danish startup aims to perform a similar trick to provide an earlier detection of someone experiencing a cardiac arrest.  The company, known as Corti, have developed an AI assistant that will analyze calls to ambulance dispatchers in Denmark to listen out for clues that point to the caller suffering a heart attack.

Being able to understand whether someone is having a heart attack might sound something that would be fairly straight forward for both caller and dispatcher, but apparently not.  Often the caller can’t accurately diagnose their condition until it’s too late, and survival chances decrease by 10% with each minute spent outside of the emergency room.

Early diagnosis

It’s estimated that dispatchers in Copenhagen can recognize cardiac arrest from symptoms given over the phone with around 73% accuracy.  The AI assistant however can achieve results of around 95%. Whilst these results were achieved in a relatively small-scale study, the team are confident they’ll be replicated in a larger study involving some 170,000 different calls, the results of which are due to be published shortly.

The system looks particularly for non-verbal cues, and is especially adroit at filtering out background noises to get to the important stuff.  For instance, in one particular call, a man had fallen from his roof and whilst the dispatcher was taking note of signs of a broken back from the fall, Corti had picked out the man gasping for breath and correctly hypothesized that he had fallen as a result of a heart attack.  Sadly, in this instance, the system was still undergoing testing so such an early intervention was not made and the man was sadly not seen to in time.

The team hope however that such early diagnoses will enable dispatchers to coach the caller through providing CPR or to better prepare things for the first responder to arrive.

The aim isn’t to replace the dispatch staff but rather to provide them with the tools to make smarter and more efficient decisions.  In a role where seconds matter and it’s crucial that the right choices are made, any help on offer must surely be welcomed.  It’s a nice example of how new AI technology can augment roles rather than replace them, and you can learn more about Corti via the video below.