Using AI to help predict cardiac arrests
Penn Today • • health
Key Points:
- Researchers at the University of Pennsylvania have developed CAMEL, an AI model that analyzes extended ECG data as a language to predict cardiac events like arrhythmias and cardiac arrest 10 to 15 minutes before they occur.
- CAMEL differs from traditional AI models by processing hours of continuous telemetry data rather than short ECG snapshots, allowing it to detect subtle, early warning signs of cardiac deterioration that conventional tools might miss.
- The interdisciplinary team combines clinical expertise from cardiologists with computer scientists' skills in pattern recognition to improve prediction accuracy and integrate ECG data with clinical notes for comprehensive analysis.
- Before clinical deployment, the model will be tested in real-time without alerting medical staff to minimize false alarms and ensure reliability, with the goal of surpassing current standards of care.
- Researchers are also exploring CAMEL's potential application in consumer wearable devices to extend cardiac risk monitoring beyond hospital settings.
