Arrhythmias & electrophysiology
Deep learning to predict left ventricular hypertrophy from the electrocardiogram
Authors' conclusion
Our findings support the feasibility of scalable DL-based screening tools for the prediction of LVH from the ECG, whilst highlighting the need for model development using larger datasets with greater diversity to ensure generalizability.
Study context
- Population
- UK Biobank imaging study participants with 12-lead electrocardiogram (ECG); external evaluation in Study of Health in Pomerania (SHIP)UK Biobank
- Participants
- 48,835 participants
- Comparison
- Deep learning (DL) model for prediction of cardiac magnetic resonance (CMR)-derived left ventricular hypertrophy (LVH)
- Outcome
- Classification of CMR-derived LVH (area under the receiving operating curve (AUROC))
Methodological detailsexternal evaluation
- External evaluation
Performance
As reported in the abstract
CMR-derived LVH, UK Biobank
| Test or model | AUC |
|---|---|
| DL model | 0.97 |
DL model
AUC0.97
- Participants with CMR-derived LVH
- 717 (1.5%)
Further findings
| Outcome | Values |
|---|---|
| External evaluation in SHIP | DL model AUC 0.78 |
- External evaluation in SHIP — Participants in SHIP
- 1423
- External evaluation in SHIP: modest generalizability; differences in clinical profile, ECG acquisition and CMR labelling
Most predictive ECG components
- QRS complex and ventricular rate most predictive of LVH
Original publication
Deep learning to predict left ventricular hypertrophy from the electrocardiogram. Europace. Published 23 January 2026. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →