Deep-learning prediction of cardiovascular outcomes from routine retinal images in individuals with type 2 diabetes
2025

Using AI to Predict Heart Problems from Eye Images in Diabetics

Sample size: 6127 publication 10 minutes Evidence: high

Author Information

Author(s): Syed Mohammad Ghouse, Trucco Emanuele, Mookiah Muthu R. K., Lang Chim C., McCrimmon Rory J., Palmer Colin N. A., Pearson Ewan R., Doney Alex S. F., Mordi Ify R.

Primary Institution: University of Dundee

Hypothesis

Can a deep-learning AI model predict cardiovascular disease outcomes from diabetic retinal images?

Conclusion

The study found that a deep-learning AI model can accurately predict major adverse cardiovascular events from routine retinal screening photographs.

Supporting Evidence

  • The AI model showed a strong correlation with traditional cardiovascular risk scores.
  • Higher retinal-predicted risk was significantly associated with increased 10-year risk of major adverse cardiovascular events.
  • The model's performance was comparable to traditional clinical risk assessments.
  • Combining AI-derived predictions with genetic risk scores improved overall risk prediction.

Takeaway

Doctors can use pictures of your eyes to help figure out if you might have heart problems, thanks to smart computer programs.

Methodology

The study used a deep-learning model to analyze retinal images from 6127 individuals with type 2 diabetes, predicting cardiovascular risk and comparing it to traditional risk scores.

Potential Biases

The study relied on electronic health records, which may not capture all relevant clinical data uniformly.

Limitations

The study primarily included individuals aged 60-75 and was predominantly Caucasian, which may limit generalizability.

Participant Demographics

Mean age was 67 years, with 55% male participants.

Statistical Information

P-Value

p<0.001

Confidence Interval

95% CI 1.04–1.06

Statistical Significance

p<0.001

Digital Object Identifier (DOI)

10.1186/s12933-024-02564-w

Want to read the original?

Access the complete publication on the publisher's website

View Original Publication