How a startup predicts heart failure over phone calls using AI and speech recognition

• Corti.ai developed technology to save thousands of people from suffering cardiac arrests.
• It combines advanced voice and pattern recognition tech to predict cardiac arrest cases in real-time.
• The technology will enable first-aid responders to provide better-informed treatment.
• See more stories on Insider's business page.

Danish startup Corti uses artificial intelligence (AI) to detect cardiac arrests.

The AI from Corti acts as a digital assistant for dispatchers taking emergency calls. The software uses automatic speech recognition to analyze the call in situations where every minute counts.

Based on data from millions of previous calls, the AI looks for signs of cardiac arrest, including both verbal and non-verbal data - like tone of voice and breathing patterns. During the call, the software provides the dispatcher with suggestions for questions and recommendations for action.

Corti is already deployed in Copenhagen, where the AI has made a clear difference.

When researchers at the emergency department tested the technology on 161,000 emergency calls from the Danish capital in 2014, Corti was 93% accurate in identifying cardiac arrest. Actual human dispatchers only got 73% right.

AI has many applications, but health care is the most obvious for Corti founder and CEO Andreas Cleve.

"It is absurd that placing an ad for a beanbag chair on a social network leverages some of the most advanced AI-tech in history, while health care professionals making life-or-death decisions have to make do with technology from the 1990s," he said. "We want to change all that and use AI where it can really make a tangible difference."

Another reason for Corti's focus on the medical field is the availability of data.

"The pre-hospital sector is very good at documenting work which means we have sufficient historical data to make the correct analyses," Andreas Cleve explained.

Corti has some prominent backers in the AI community, amongst others Danny Lange, an investor in Nordic venture capital firm ByFounders.

The 55-year-old Dane has developed machine learning and AI applications for some of the prominent tech companies in the world, including Amazon, Google, Microsoft, and Uber.

"Corti is an excellent example of intelligent use of AI in real life. It actually saves lives." Danny Lange said. "By improving the human decision process in challenging situations. That is applicable in many more areas, such as ER doctors who have to make life-and-death decisions in minutes."

The Danish AI veteran expects many more practical AI applications in the coming years.

"AI has been in use for more than a decade, but it was mainly in the large tech companies with their in-house resources. Now, AI is going mainstream and being democratized," Lange added. "Any company - not just Google, Amazon, IBM, and Microsoft - can now get AI resources through a platform, like any other web service."

Corti's software analyzes dialogue in difficult environments

In Corti's case, however, the entire solution has been developed in-house. "We decided to build our own models for speech recognition from the ground up, to suit patient-doctor conversations." Andreas Cleve explained.

"The technology you find in applications like Alexa [Amazon's voice assistant] is different because it's built for user prompts, monologues, and short sentences," he said. "We wanted to build a solution that would be a part of a dialogue in a very difficult acoustic environment where even the background noise might be important."

Besides the investment from ByFounders, Corti has previously received funding from venture fund Sunstone.