Clues to mortality risk found in the retina

Link made between mortality risk and the gap between what AI predicts an individual’s age to be and their actual age

green eye
Pixabay/zehra soy
Scientists have developed a deep learning model that predicts an individual’s age on the basis of fundus images.

The research, which is published in British Journal of Ophthalmology, involved using more than 80,000 fundus images to train and validate a deep learning model for age prediction.

On average, the model predicted an individual’s age to within 3.55 years.

By analysing mortality data, the scientists found that each one-year increase in the gap between actual age and predicted age was associated with a 2% increase in all-cause mortality.

“Our findings indicate that retinal age gap might be a potential biomarker of ageing that is closely related to risk of mortality, implying the potential of retinal image as a screening tool for risk stratification and delivery of tailored interventions,” the researchers highlighted.

The authors concluded that the deep learning can detect “footprints of ageing in fundus images” and predict age with high accuracy.

“Our work calls for future research into applications of the retinal age gap, and whether retinal age can be used to better understand processes underpinning ageing,” the authors added.