Cross-sectional study
Automated Assessment of Cardiac Systolic Function From Coronary Angiograms With Video-Based Artificial Intelligence Algorithms
- Cross-sectional study
- external validation
Authors' conclusion
This cross-sectional study represents an early demonstration of estimating LVEF from standard angiogram videos of the left coronary artery using video-based DNNs.
Study context
- Population
- Adults with a coronary angiogram and TTE within 3 months before or 1 month after it, University of California, San Francisco (UCSF) 2012–2019; external validation at University of Ottawa Heart Institute4,042 angiograms
- Participants
- 3,679 adults
Main finding
As reported in the abstract
Reduced LVEF (≤40%) discrimination and continuous LVEF prediction vs TTE
| Group | AUROC, UCSF test (n = 813) | Diagnostic OR for reduced left ventricular ejection fraction (LVEF), UCSF test | Mean absolute error (MAE) vs TTE LVEF, UCSF test | Difference ≤5% vs TTE LVEF, UCSF test | Difference >15% vs TTE LVEF, UCSF test | AUROC, external validation (n = 776) | MAE, external validation |
|---|---|---|---|---|---|---|---|
| CathEF | 0.911 (95% CI, 0.887-0.934) | 22.7 (95% CI, 14.0-37.0) | 8.5% (95% CI, 8.1%-9.0%) | 38.0% (309 of 813) | 15.2% (124 of 813) | 0.906 (95% CI, 0.881-0.931) | 7.0% (95% CI, 6.6%-7.4%) |
Observational data: associations, not effects.
Further findings
Performance by subgroup (sex, BMI, eGFR ≤45, ACS, obstructive CAD, LV hypertrophy)
- Performance consistent across subgroups
Original publication
Automated Assessment of Cardiac Systolic Function From Coronary Angiograms With Video-Based Artificial Intelligence Algorithms. JAMA Cardiol. Published 1 June 2023. doi:10.1001/jamacardio.2023.0968
Summarised from the published abstract; every number is checked against it