Heart failure & cardiomyopathiesDiagnostic or prognostic study
Deep Learning-Based Prediction of Right Ventricular Ejection Fraction Using 2D Echocardiograms
- Diagnostic accuracy study
- retrospective
- internal and external validation
Authors' conclusion
Using 2D echocardiographic videos alone, the proposed DL-based tool can accurately assess RV function, with similar diagnostic and prognostic power as 3D imaging.
Study context
- Population
- Patients with right ventricular ejection fraction (RVEF) measured by 3D echocardiographyexternal set 365
- Participants
- 831 patients
- Comparison
- Ensemble model vs Expert reader
- Outcome
- Accuracy of DL-predicted RVEF vs 3D echocardiography-derived RVEF
Performance
As reported in the abstract
RVEF prediction, mean absolute error
| Test or model | Mean absolute error, internal validation | Mean absolute error, external validation |
|---|---|---|
| Ensemble model | 4.57 percentage points | 5.54 percentage points |
Further findings
| Outcome | HR (95% CI) | P value | |
|---|---|---|---|
| MACE, by DL-predicted RVEF (adjusted for age, sex, LV systolic function) · median follow-up 1.9 years | 0.924 (0.862–0.990) | P = 0.025 | |
| 0.512Hazard ratio (log scale) |
| Outcome | Ensemble model | Expert reader | Estimate | P value |
|---|---|---|---|---|
| RV dysfunction (RVEF <45%), external validation | ||||
| Accuracy | Not reported in the abstract | 77.0% | P = 0.678 |
- Right ventricular (RV) dysfunction (RVEF <45%), external validation: Deep learning (DL) ensemble model Accuracy 78.4%
Original publication
Deep Learning-Based Prediction of Right Ventricular Ejection Fraction Using 2D Echocardiograms. JACC Cardiovasc Imaging. Published 10 May 2023. doi:10.1016/j.jcmg.2023.02.017
Summarised from the published abstract; every number is checked against it