A simple photograph of the back of the eye could reveal certain signs of aging that are not as easily visible elsewhere.
Researchers trained an artificial intelligence system to estimate a person’s age from their retina. To do so, they used images from 71,343 participants in the UK Biobank, a large British medical database.
Close-up of an eye.
Pixabay
The data were divided into five groups. At each stage, four groups were used for training and the fifth for testing. The process was repeated five times so that each participant was assessed by a model that had not learned from their images.
With this method, the average difference between predicted and actual age was 2.85 years. The researchers then tested the AI on an independent population from the Rotterdam Study. Its accuracy remained similar, with an average difference of about 3.35 years per person.
How does AI find age in an eye? It relies on very subtle changes that appear over the years. Blood vessels, the central region of the retina and the area around the optic nerve all play a role in its decisions.
But the study’s goal was not limited to correctly guessing a person’s date of birth. The researchers were mainly interested in cases where the retina appeared older or younger than expected.
They call this difference the “retinal age gap.” For example, a 60-year-old person whose retina appears to be 65 has a positive gap of five years.
This gap appears to contain information about the body’s overall health. A retina that appeared older was notably associated with diabetes, high blood pressure, smoking and certain signs of inflammation. Links also existed with poorer performance on certain memory tests.
The researchers then tracked the onset of diseases over time. A larger retinal age gap was associated with more cardiovascular disease, strokes, dementia and certain cancers. It was also linked to higher mortality.
This does not mean that an “old” retina indicates that a person will develop one of these diseases. These are statistical associations observed in large groups. The study also does not show that retinal aging causes these health problems.
Differences also appear between women and men. In men, the retinal gap was more strongly linked to metabolic problems, such as high blood pressure or the collection of factors associated with diabetes and cardiovascular disease.
In women, retinal blood vessels appear to play a more important role in the predictions. After menopause, the retina also appears, on average, to be slightly older. The medical associations then become more similar to those observed in men.
The team finally compared this indicator with the available genetic data. It found variations linked to longevity, metabolism, brain aging and certain eye diseases.
The next step will be to determine whether this measurement genuinely offers doctors anything beyond already known risk factors. A retinal photograph is convenient because it is quick, painless and already commonly used in ophthalmology.