Deep-learning prediction of cardiovascular outcomes from routine retinal images in individuals with type 2 diabetes
Authors' conclusion
A deep-learning AI model could accurately predict MACE from routine retinal screening photographs with a comparable performance to traditional clinical risk assessment in a diabetic cohort. Combining the AI-derived retinal risk prediction with a coronary heart disease polygenic risk score improved risk prediction. AI retinal assessment might allow a one-stop CVD risk assessment at routine retinal screening.
Study context
- Population
- Individuals with T2D without MI or stroke prior to study entry1,241 in test cohort
- Participants
- 6,127 individuals
- Comparison
- Retina-predicted cardiovascular disease (CVD) risk vs pooled cohort equation (PCE) risk score
- Outcome
- Time to first MACE:
- CV death
- MI
- Stroke
- Follow-up
- 10-year
Methodological detailsdeep-learning artificial intelligence (AI) model (EfficientNet-B2), training/validation/testing cohorts (70%/10%/20%)
- Deep-learning AI model (EfficientNet-B2)
- Training/validation/testing cohorts (70%/10%/20%)
Performance
As reported in the abstract
MACE
| Test or model | AUC |
|---|---|
| Retinal risk score | 0.697 |
| PCE | 0.697 |
AUC
Retinal risk score0.697
PCE0.697
- Retina-predicted risk and major adverse CV event (MACE) risk, HR per 1% increase
- 1.05 (95% CI 1.04-1.06, p < 0.001)
- Retina-predicted risk and MACE risk adjusted for PCE and polygenic risk score (PRS), HR
- 1.03 (95% CI 1.02-1.04, p < 0.001)
- MACE events
- 288
Further findings
Combined retinal risk score + PCE + PRS vs PCE alone
- AUC of combined model; significantly improved vs PCE alone
- 0.728
Correlation of retina-predicted CVD risk
- Correlation with PCE risk score, r
- 0.66
- Correlation with PRS, r
- 0.05
Change in retina-predicted risk · within 3 years
- Increase in retina-predicted risk associated with subsequent increased MACE likelihood
Original publication
Deep-learning prediction of cardiovascular outcomes from routine retinal images in individuals with type 2 diabetes. Cardiovasc Diabetol. Published 2 January 2025. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →