Valvular & structural heart diseaseDiagnostic or prognostic study
A Deep Learning Model to Identify Mitral Valve Prolapse From the Echocardiogram
- Diagnostic accuracy study
- deep learning model
- external validation
Authors' conclusion
A deep learning model identifies MVP from echocardiogram videos, and model predictions are associated with clinical endpoints including MR and future MVR.
Study context
- Population
- Cardiology and primary care patients with transthoracic echocardiography at MGH and BWHderivation sample
- Participants
- 16,902 participants
- Comparison
- MGH cardiology internal validation vs MGH primary care external validation vs BWH primary care external validation
- MGH primary care external validationn = 8,888
- BWH primary care external validationn = 257
- Outcome
- Identification of MVP from echocardiogram videos by DROID-MVP
Performance
As reported in the abstract
MVP identification by DROID-MVP
| Test or model | AUROC | Average precision | Prevalence |
|---|---|---|---|
| MGH cardiology internal validation | 0.947 (95% CI: 0.910-0.984) | 0.682 (95% CI: 0.565-0.784) | 0.036 |
| MGH primary care external validation | 0.964 (95% CI: 0.951-0.977) | 0.651 (95% CI: 0.578-0.716) | 0.022 |
| BWH primary care external validation | 0.968 (95% CI: 0.946-0.989) | 0.774 (95% CI: 0.666-0.797) | 0.113 |
Further findings
| Outcome | Estimate | P value | |
|---|---|---|---|
| Moderate or severe MR, high (>0.67) vs low (<0.33) DROID-MVP score | adjusted OR 2.0 (95% CI 1.1–3.8) | P = 0.030 | |
| Future MVR, high (>0.67) vs low (<0.33) DROID-MVP score | adjusted HR 3.7 (95% CI 1.5–8.9) | P = 0.004 | |
| 0.10.31310Ratio (log scale) |
Mitral valve prolapse (MVP) prevalence, derivation sample
- Patients with MVP
- 783 (4.6%)
Original publication
A Deep Learning Model to Identify Mitral Valve Prolapse From the Echocardiogram. JACC Cardiovasc Imaging. Published 30 September 2025. doi:10.1016/j.jcmg.2025.08.011
Summarised from the published abstract; every number is checked against it