VF-FM
Foundation Model for Defibrillation Strategy Prediction
Building the first foundation model to predict optimal defibrillation strategy from ECG waveforms during CPR.
The Problem
Over 60% of ventricular fibrillation (VF) cases remain refractory to initial defibrillation shocks. Current guidelines follow a one-size-fits-all approach, potentially delaying escalation to advanced defibrillation strategies like vector change (VC) or double sequential external defibrillation (DSED).
Our Approach
We're developing a foundation model using self-supervised learning on large ECG datasets, then fine-tuning for defibrillation outcome prediction. The model is designed to work with artifacted signals during chest compressions, enabling real-time decision support.
Innovation
First AI system to predict optimal shock strategy (standard dose vs vector change vs DSED) in real-time during CPR, accounting for compression artifacts that corrupt traditional ECG analysis.
Project Timeline
- Lawrence Leroux (PI)
- Sheldon Cheskes (Co-PI)
- Jamal Bentahar
- Lyes Kadem
- Yiorgos Alexandros Cavayas
RSN Funding - $40,000 CAD (2024-2026)
ZOLL Foundation - $36,000 USD (2025-2026)
Ethics approval obtained