Heart failure & cardiomyopathies
Improved diagnosis of arrhythmogenic right ventricular cardiomyopathy using electrocardiographic deep learning
Authors' conclusion
ECG-DL augments diagnosis of ARVC to the level of an ARVC expert and can differentiate true ARVC diagnosis from phenotype-mimics and at-risk family members/genotype-positive individuals.
Study context
- Population
- Patients referred for arrhythmogenic right ventricular cardiomyopathy (ARVC) evaluation at Johns Hopkins Hospital; external validation in carriers of P/LP ARVC gene variants (Geisinger MyCode)external cohort N = 167
- Participants
- 688 patients
- Comparison
- Electrocardiogram (ECG) deep learning model vs ECG interpretation by referring cardiologist and ARVC expert
- Outcome
- Discrimination (c-statistic) for TFC-defined ARVC diagnosis
Methodological detailsdevelopment and external validation
- Development and external validation
Performance
As reported in the abstract
Discrimination for ARVC diagnosis (JHH hold-out testing cohort)
| Test or model | C-statistic |
|---|---|
| ECG-DL | 0.87; 0.86-0.89 |
| Referring cardiologist | 0.53; CI 0.52-0.53 |
| ARVC expert | 0.85; 0.84-0.86 |
C-statistic
ECG-DL0.87; 0.86-0.89
Referring cardiologist0.53; CI 0.52-0.53
ARVC expert0.85; 0.84-0.86
Further findings
- ECG-DL discrimination, Geisinger external validation cohort
- 0.80; 0.77-0.83 · c-statistic
- ECG-DL combined with non-ECG-derived Task Force Criteria (TFC)
- 0.940; 95% CI 0.933-0.948 · c-statistic, Johns Hopkins Hospital (JHH) testing · c-statistic, Geisinger validation 0.897; 95% CI 0.883-0.912
- ECG-DL discrimination, Geisinger external validation cohort: ARVC prevalence n = 17 [10.2%]
Original publication
Improved diagnosis of arrhythmogenic right ventricular cardiomyopathy using electrocardiographic deep learning. Heart Rhythm. Published 20 August 2024. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →