Electrocardiogram-based deep learning improves outcome prediction following cardiac resynchronization therapy
Problem
Current guideline ECG criteria poorly predict which patients will benefit from cardiac resynchronization therapy (CRT).
Why it matters
Poor CRT patient selection leads to death, need for left ventricular assist device, or heart transplantation.
What the AI did
A deep learning algorithm trained on 1.1 million ECGs compressed ECGs into 21 explainable factors (FactorECG) that predicted CRT outcomes better than existing criteria.
Results
c-statistic 0.69 (95% CI 0.66-0.72) vs QRSAREA 0.61 and guideline criteria 0.57, P<0.001 for both.
Impact
Deep learning significantly outperformed existing ECG criteria in a multi-centre cohort of 1306 CRT patients.
Record
| Venue | European Heart Journal |
|---|---|
| Published | 2022 |
| Work type | article |
| Authors | Philippe C. Wouters, Rutger R van de Leur, Melle B. Vessies, Antonius M W van Stipdonk, Mohammed Ali Ghossein, Rutger J. Hassink, Pieter A. Doevendans, Pim van der Harst et al. |
| Citations | 57 (OpenAlex, at import time) |
| References | 37 |
| Model or method | FactorECG (deep learning-based; architecture not stated in source) |
| Organization | not stated in source |
| Primary source | https://doi.org/10.1093/eurheartj/ehac617 |
| Open access | hybrid |
| Keywords | ECG, deep learning, cardiac resynchronization therapy, outcome prediction, explainable AI |
Importance rating
significant — thousands to millions affected
Poor CRT selection leads to death, LVAD, or transplantation; deep learning on 1.1M ECGs significantly outperformed guideline criteria in multi-centre cohort.