Project Overview
AI software that listens to emergency calls to help dispatchers identify signs of cardiac arrest quickly and accurately.
Layman's Explanation
Corti’s AI system listens to emergency calls to detect verbal and non-verbal cues that may indicate a cardiac arrest, helping dispatchers decide faster on life-saving measures.
Analogy
In healthcare, this AI acts like a digital assistant in a hospital's triage, recognizing the urgency of a patient’s symptoms and prompting the team to respond immediately.
Details
Corti's software is designed to identify out-of-hospital cardiac arrests during emergency calls by recognizing specific speech patterns, caller tone, and breathing cues. Tested in Copenhagen, it achieved a 93.1% accuracy rate, surpassing the 72.9% rate by human dispatchers, and could detect cardiac arrest indicators within 48 seconds on average—substantially faster than human response time. The AI guides dispatchers by suggesting follow-up questions and advising on CPR instructions if needed, but the final decision remains with human staff. Despite its advantages, the software has faced scrutiny for its opaque decision-making process, a common challenge for complex machine learning models in healthcare.