Corti, the Copenhagen-based artificial intelligence platform that uses machine learning to provide emergency medical personnel with life-saving diagnostic assistance, has announced a partnership with the European Emergency Number Association (EENA), with four sites across Europe being selected to pilot the technology. The project has the potential to save thousands of lives, and change forever the way emergency medical calls are handled.
The four sites will be chosen by June 2018, and will start running the platform late summer/early September 2018 through to April 2019.
Corti is already being used by Copenhagen Emergency Medical Services (EMS) to detect cardiac arrests during emergency calls. A new study shows that Corti is 20.2% more accurate at detecting Out of Hospital Cardiac Arrests (OHCA) than medical dispatchers, and is also significantly faster (31 seconds) at doing so.
161,650 emergency calls (158,330 non-OHCA and 2,157 OHCA calls) were analysed in 2014 using Corti’s machine learning platform. The model reached a detection rate (sensitivity) of 93.1%, and a specificity (detection of those without the condition) rate of 98%. In comparison, human dispatchers achieved a sensitivity of 72.9%, meaning that the Corti platform is 20.2% more accurate in terms of detecting cardiac arrests than human dispatchers.
Corti’s time-to-recognition is 48 seconds, compared to 1 minute 19 seconds for the human dispatchers.
How Corti works
- Corti is built on state-of-the-art, real-time automatic speech recognition technology that can act as a digital assistant for medical dispatchers handling emergency calls, detecting out-of-hospital cardiac arrests
- Deep neural networks process the conversation in real-time, looking for verbal and non-verbal patterns of communication – such as tone of voice and breathing issues – that might not be obvious to a human, and even when there is heavy background noise
- Using this data, Corti provides human call-takers with recommendations for how to best triage patients, resulting in significantly reduced error rates, faster critical diagnoses, and lives being saved
When the model at each pilot site is trained to a certain level, the site team will be able to test it with their local personnel. After 6 months of testing, Corti, EENA, and the four pilot sites will present a report at the EENA Conference 2019, detailing their findings.
