Arrhythmias & electrophysiologyDiagnostic or prognostic study
Deep Learning–Based Continuous QT Monitoring to Identify High-Risk Prolongation Events After Class III Antiarrhythmic Initiation
- Diagnostic accuracy study
- external validation
- deep learning model development
Authors' conclusion
A single-lead, deep-learning approach can achieve guideline-level measurement accuracy, enable continuous QTc surveillance from nonstandard ECG vectors, and identify clinically meaningful outpatient QTc prolongation associated with a >4-fold increase in serious ventricular arrhythmias.
Study context
- Population
- Patients on class III antiarrhythmics (dofetilide or sotalol); model development on 12-lead ECGs from a large health system, external test in another centre
- Comparison
- Single-lead deep learning QT/QTc prediction vs clinician-adjudicated QT
Performance
As reported in the abstract
Classification of prolonged QTc from single-lead signals
| Test or model | AUC, internal test set | Mean absolute error, internal test set | AUC, external | Mean absolute error, external |
|---|---|---|---|---|
| 3DRECON-QT | 0.942 | 17.5 ms | 0.943 | 21.1 ms |
Further findings
Continuous dofetilide monitoring: QTc prediction vs ground truth
- Correlation (r)
- 0.851
- Mean absolute error
- 17.8 ms
- AUC for prolonged QTc
- 0.936
- AUC for ≥15% QTc rise
- 0.816
QTc prediction from true insertable cardiac monitor recordings
- Correlation (r)
- 0.824
- Mean absolute error
- 17.5 ms
Original publication
Deep Learning–Based Continuous QT Monitoring to Identify High-Risk Prolongation Events After Class III Antiarrhythmic Initiation. Circulation. Published 8 November 2025. doi:10.1161/circulationaha.125.077494
Summarised from the published abstract; every number is checked against it