Valvular & structural heart diseaseCohort study
Deep Learning Model of Diastolic Dysfunction Risk Stratifies the Progression of Early-Stage Aortic Stenosis
- Retrospective cohort study
- external validation in 2 cohorts
Authors' conclusion
Assessment of DD using DL can stratify the latent risk associated with the progression of early-stage AS.
Study context
- Population
- Participants with aortic valve sclerosis (ARIC cohort); validation in mild-to-moderate aortic stenosis (AS) (cardiac magnetic resonance (CMR) cohort) and aortic valve (AV) sclerosis (positron emission tomography (PET)/CT cohort)ARIC cohort
- Participants
- 898 participants
- Exposure
- Higher vs lower deep learning (DL)-predicted probability of diastolic dysfunction
- Outcome
- New diagnosis of AS; composite of subsequent mortality or AV interventions
Main finding
As reported in the abstract
Higher DL-predicted probability of DDAdjusted hazard ratio (95% CI)P value
development of AS
3.482(2.061–5.884)
P < 0.001mortality or AV interventions
7.033(3.036–16.290)
P < 0.001Observational data: associations, not effects.
Further findings
Prediction of AS progression by ARIC-derived multivariable Cox model (CMR cohort, n = 50)
- C-index
- 0.798 (95% CI 0.648-0.948)
Correlation of model predictions with valvular 18F-NaF mean SUV (PET/CT cohort, n = 18)
- r (P = 0.008)
- 0.62
Original publication
Deep Learning Model of Diastolic Dysfunction Risk Stratifies the Progression of Early-Stage Aortic Stenosis. JACC Cardiovasc Imaging. Published 18 September 2024. doi:10.1016/j.jcmg.2024.07.017
Summarised from the published abstract; every number is checked against it