AgingResearch.NewsJoin free
AHA Scientific Sessions 2026 Abstract

Adjusted eye-image model stayed accurate in Korean cohort

In 12,299 Korean screening participants, the authors reported that adjusted probabilities matched observed artery-calcium frequencies to approximately within one percentage point.

Paper by Dongjin Nam, Sahil Thakur, Simon Nusinovici, +8 more authorsByung Hyun Joe · Yonsei University · Mediwhale
Retina with branching vessels beside a small group of people and artery cross-sections showing sparse calcium with a wide opening and dense calcium narrowing…

Our read

Biomarkers
Studied:
Cohort study of 12,299 Korean screening participants (Independent replication, multicenter)
What changed:
Adjusted artery-calcium predictions matched observed frequencies to approximately within one percentage point in the independent Korean cohort
The catch:
The analysis measured prediction accuracy, not whether model use improved health outcomes
Why it’s aging research:
It bears on staying healthy in midlife and later life.

Editor of record: Boris Djordjevic · AgingResearch.News

Paper
Follow @agingbrief

Korean screening participants and people referred for a heart scan in Greece were included in an observational validation of an eye-image prediction model. The team adjusted the model using 1,834 Korean participants, then tested it in 12,299 other Korean participants and 1,195 people in Greece. The model used images of the retina, the light-sensitive tissue at the back of the eye, to estimate the probability of calcium buildup in heart arteries.

Adjusted predictions were accurate overall and locally in the independent Korean cohort. The model also retained its ability to distinguish people above and below calcium thresholds used to guide preventive care. Performance was weaker in the Greek cohort, particularly at the highest calcium threshold. The authors concluded that the results supported using the model to help select people for heart scans.

Why it matters

Calcium buildup in heart arteries helps guide prevention of coronary artery disease, an age-related illness. The work addresses whether eye images could help prioritize heart scans where access is uneven.

Caveats

The analysis assessed prediction accuracy and estimated decision-making benefit, not whether using the model improved health outcomes. The findings were reported in a conference abstract, and European validation involved one Greek referral center.

Comments

Be the first to comment.

More on Deep learning