Diagnostic or prognostic study
Artificial Intelligence Automation of Echocardiographic Measurements
- Diagnostic accuracy study
- development and external validation
- temporal split
Authors' conclusion
EchoNet-Measurements achieves high accuracy in automated echocardiographic quantification and potential for assisting the clinicians in the echocardiography workflow.
Study context
- Population
- Echocardiography studies from Cedars-Sinai Medical Center (2011–2023), with external validation at Stanford Healthcarestudies
- Participants
- 155,215 studies
- Comparison
- Deep learning automated measurement vs sonographer measurement
Performance
As reported in the abstract
Accuracy vs sonographer measurements, all 9 B-mode and 9 Doppler measurements
Mean coverage probability, held-out Cedars-Sinai Medical Center (CSMC) test set
0.796
Mean coverage probability, external Stanford Healthcare (SHC) data set
0.839
Mean relative difference, held-out CSMC test set
0.120
Mean relative difference, external SHC data set
0.096
Further findings
| Outcome | EchoNet-Measurements | Sonographer |
|---|---|---|
| End-to-end evaluation on 2,103 temporally distinct CSMC studies | ||
| Mean coverage probability | 0.803 | Not reported in the abstract |
| Mean relative difference | 0.108 | Not reported in the abstract |
Performance by age, sex, AF, obesity status and machine vendor
- Performance consistent across patient characteristics and machine vendors
Original publication
Artificial Intelligence Automation of Echocardiographic Measurements. J Am Coll Cardiol. Published 7 September 2025. doi:10.1016/j.jacc.2025.07.053
Summarised from the published abstract; every number is checked against it