AI Case Study

Danish emergency service dispatchers identify heart-attacks in real-time emergency calls with 95% accuracy, compared to 73% for human dispatchers, with real-time speech analysis and machine learning

Danish Emergency Service dispatchers are assisted by real-time models that can predict with 95% accuracy, regardless of who is calling or the background noise, whether someone is having a heart attack. This compares to 73% accuracy for human dispatchers who have extensive training. The models have been trained on millions of hours of call.

Industry

Public And Social Sector

Public Services

Project Overview

"In Copenhagen, dispatchers now have help from AI. If you call for an ambulance, an artificially intelligent assistant called Corti will be on the line, using speech recognition software to transcribe the conversation, and using machine learning to analyze the words and other signals in the background that point to a heart attack diagnosis. The dispatcher gets alerts from the bot in real time."

“When an emergency call is made, the dispatcher using Corti will be able to utilize what is essentially a digital assistant that listens in on the conversation,” explains Cleve. “Deep neural networks process the conversation in real-time, looking for verbal and non-verbal patterns of communication, including tone of voice, breathing difficulties, and other metadata, that humans might not be immediately obvious to a human, particularly in the heat of the moment.”

"During the call, the data is analyzed and compared with historical data collected from millions of previous emergency calls, which is how the platform learns over time. As understanding of the incident increases during the call, achieved both through the analyzed data and the caller’s input, Corti will learn to predict the criticality of the situation, delivering real-time alerts and recommendations to the dispatcher."

"At the core of Corti’s AI is a state-of-the-art, real-time automatic speech recognition technology that provides recommendations and advice for medical dispatchers handling emergency calls."

Reported Results

"In an early, small-scale study, the machine learning model knew the calls were reporting cardiac arrest 95% of the time." This is versus "dispatchers...ho are well trained, can recognize cardiac arrest from descriptions over the phone around 73% of the time."

Technology

"At the core of Corti’s AI is a state-of-the-art, real-time automatic speech recognition technology that provides recommendations and advice for medical dispatchers handling emergency calls."

“Deep neural networks process the conversation in real-time, looking for verbal and non-verbal patterns of communication, including tone of voice, breathing difficulties, and other metadata, that humans might not be immediately obvious to a human, particularly in the heat of the moment.”

Function

Operations

General Operations

Background

"When someone goes into cardiac arrest outside of a hospital, time is critical. The chance of survival decreases about 10% with each minute. The first step–recognizing that it’s cardiac arrest, when your heart fully stops–is challenging for emergency dispatchers on the phone, who have to make sense of symptoms relayed by a panicked friend or relative."

Danish Emergency Services wanted to know if they could use AI to improve survival rates by better analysing and identifying cardiac arrest situations during calls to dispatchers. "During emergencies there can be chaos and a myriad of factors that are hard to control, such as noise and other disruptions." Could they better predict cardiac arrest regardless of whether it was the patient or another making the call? Could they better predict when there was a lot of background noise?

Benefits

Data

Machine learning models have "been trained to detect the signs of a heart attack through analysis of millions of calls."

"During the call, the data is analyzed and compared with historical data collected from millions of previous emergency calls, which is how the platform learns over time:"