Heart failure & cardiomyopathiesCohort study
Deep Learning–Enabled Assessment of Left Heart Structure and Function Predicts Cardiovascular Outcomes
- Retrospective cohort study
- multi-institutional
- external validation
Authors' conclusion
Deep learning echocardiographic interpretation accurately quantified standard measures of left heart structure and function, which in turn were associated with future clinical outcomes.
Study context
- Population
- Ambulatory cardiology and primary care patients with echocardiograms64,028 echocardiograms (training)
- Participants
- 27,135 participants
Main finding
As reported in the abstract
Deep learning view classification and left heart quantification; association with incident outcomes
View discrimination (PLAX, apical 4- and 2-chamber) vs human expert annotation, AUC
>0.97
Standard left heart measures vs study report values, R2 range
0.53 to 0.91
1-SD lower model-LVEF and HF, HR
1.43 (95% CI 1.23-1.66)
1-SD lower model-LVEF and death, HR
1.17 (95% CI 1.06-1.30)
Observational data: associations, not effects.
Further findings
Other model-derived left heart measures
- Model-derived measures predicted incident HF, AF, MI and death; similar results for other measures
Original publication
Deep Learning–Enabled Assessment of Left Heart Structure and Function Predicts Cardiovascular Outcomes. J Am Coll Cardiol. Published 14 November 2023. doi:10.1016/j.jacc.2023.09.800
Summarised from the published abstract; every number is checked against it