Arrhythmias & electrophysiologyDiagnostic or prognostic study
Deep Learning–Augmented Electrocardiography (ECG) Analysis for Screening and Genotype Prediction of Congenital Long QT Syndrome
- Diagnostic accuracy study
- multicentre
- external validation
- cross-sectional validation
Authors' conclusion
The deep learning model improved detection of congenital LQTS from resting ECGs and allowed for differentiation between the 2 most common genetic subtypes.
Study context
- Population
- Patients with suspected inherited arrhythmia (HiRO Registry, 2012–2021); Long QT syndrome (LQTS) cohort = probands and relatives with pathogenic or likely pathogenic KCNQ1 or KCNH2 variants
- Participants
- 990 patients · 4,521 ECGs
- Comparison
- Convolutional neural network (CNN)-based ECG analysis vs QTc-based detection
- Outcome
- AUC, F1 score and sensitivity for detecting LQTS and differentiating genotypes (LQTS1 vs LQTS2)
Performance
As reported in the abstract
LQTS detection, external validation (101 patients)
| Test or model | AUC | F1 score | Sensitivity | AUC, genotype differentiation |
|---|---|---|---|---|
| CNN | 0.93 (95% CI, 0.89-0.96) | 0.84 (95% CI, 0.78-0.90) | 0.90 (95% CI, 0.86-0.94) | 0.91 (95% CI, 0.86-0.96) |
| QTc | 0.22 (95% CI, 0.13-0.31) | 0.36 (95% CI, 0.23-0.47) |
Further findings
| Outcome | CNN | QTc |
|---|---|---|
| LQTS detection with normal or borderline QTc (subgroup) | ||
| F1 score | 0.70 (95% CI, 0.40-1.00) | Not reported in the abstract |
| Sensitivity | 0.78 (95% CI, 0.53-0.95) | Not reported in the abstract |
| LQTS detection, cross-sectional validation cohort (406 patients) | ||
| AUC | 0.81 (95% CI, 0.80-0.85) | 0.74 (95% CI, 0.69-0.78) |
Original publication
Deep Learning–Augmented ECG Analysis for Screening and Genotype Prediction of Congenital Long QT Syndrome. JAMA Cardiol. Published 1 April 2024. doi:10.1001/jamacardio.2024.0039
Summarised from the published abstract; every number is checked against it