Retinal imaging for the assessment of stroke risk: a systematic review.
Level 1 - systematic review of randomized trials
Systematic review of prospective studies evaluating prognostic risk factors
PubMed 38430271 · doi:10.1007/s00415-023-12171-6
What was done
A systematic review of prospective studies was performed to evaluate retinal imaging-derived biomarkers and predictive models as indicators of future stroke risk.
What was found
Twenty-four prospective studies were identified. Wider retinal venules, lower fractal dimension, increased arteriolar tortuosity, retinopathy, and retinal emboli were associated with higher stroke risk; narrower arterioles, microaneurysms, and arteriovenous nicking showed weaker evidence. Three artificial intelligence models predicted stroke comparably to conventional risk scores but did not significantly outperform them. The abstract reports no numerical effect sizes or predictive accuracy metrics.
Why it matters
While retinal imaging non-invasively captures microvascular changes linked to stroke, current retinal-based models do not yet improve on routine clinical risk assessment tools.
Limits
The abstract lacks quantitative effect estimates (hazard ratios, odds ratios, or area under the curve) and meta-analytic pooling. Only two included studies utilized optical coherence tomography (OCT).
Cited by
- supports AI analysis of retinal images can predict coronary artery calcium scores, stroke risk, and heart disease.