Valvular & structural heart disease
Identifying Mitral Valve Prolapse at Risk for Arrhythmias and Fibrosis From Electrocardiograms Using Deep Learning
Authors' conclusion
CNN-analyzed 12-lead ECGs can detect MVP at risk for ventricular arrhythmias, death and/or fibrosis and can identify novel ECG correlates of arrhythmic risk.
Study context
- Population
- Patients with mitral valve prolapse, University of California-San Francisco, 2012–202087 of mitral valve prolapse (MVP) patients had contrast cardiac magnetic resonance (CMR)
- Participants
- 569 vs 87
- Comparison
- Convolutional neural network (CNN)-analysed 12-lead electrocardiogram (ECG) for detection of complex ventricular ectopy (ComVE), mortality and late gadolinium enhancement (LGE)
- CNN (ComVE)n = 569
- CNN (LGE)n = 87
- Outcome
- CNN detection of complex ventricular ectopy (ComVE), death and/or myocardial fibrosis (LGE) on CMR
Methodological detailsdeep learning (CNN), single-centre
- Deep learning (CNN)
- Single-centre
Performance
As reported in the abstract
Detection of ComVE
| Test or model | AUC |
|---|---|
| CNN (ComVE) | 0.80 (95% CI: 0.77-0.83) |
CNN (ComVE)
AUC0.80 (95% CI: 0.77-0.83)
- Prevalence of ComVE
- 28% (160/569)
Further findings
ComVE detection AUC, excluding moderate-severe mitral regurgitation
- AUC of CNN for ComVE, excluding moderate-severe mitral regurgitation
- 0.80 (95% CI: 0.77-0.83)
ComVE detection AUC, excluding bileaflet MVP
- AUC of CNN for ComVE, excluding bileaflet MVP
- 0.81 (95% CI: 0.76-0.85)
All-cause mortality detection
- AUC to detect all-cause mortality
- 0.82 (95% CI: 0.77-0.87)
LGE (papillary muscles or basal inferolateral wall) on CMR
- LGE prevalence among patients with available CMR
- 24%
- AUC for detection of LGE
- 0.75 (95% CI: 0.68-0.82)
ECG segments relevant to ComVE prediction
- Related to ventricular depolarization/repolarization (early-mid ST-segment and QRS from V1, V3, and III)
Original publication
Identifying Mitral Valve Prolapse at Risk for Arrhythmias and Fibrosis From Electrocardiograms Using Deep Learning. JACC Adv. Published 5 August 2023. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →