Researchers trained AI to predict human lifespan from the eyes

The gap between the biological age of the retina — the light-sensitive layers of nerve tissue at the back of the eye — and a person's actual (chronological) age is linked to their risk of death. This is stated in a study published in the British Journal of Ophthalmology.

The scientists aimed to develop a deep learning model that predicts age from an image of the fundus of the eye (retinal age), and to study the link between the retinal age gap (retinal age predicted by the deep learning model minus chronological age) and mortality risk. 

The researchers suggest that this "retinal age gap" could be used as a screening tool. A growing body of data suggests that the network of small blood vessels (the microcirculation) in the retina may be a reliable indicator of the overall state of the circulatory system and brain.

Although the risk of disease and death increases with age, it is clear that these risks vary considerably among people of the same age, meaning that "biological aging" is unique to each person and may be a better indicator of current and future health, the researchers believe.

Several tissue, cellular, chemical and imaging indicators have been developed to determine biological aging, which does not match chronological aging. However, these methods raise ethical and privacy issues and are often invasive, expensive and labor-intensive.

Therefore, the researchers turned to deep learning to see whether it can accurately predict the age of the human retina from fundus images, the inner back surface of the eye, and to see whether there is any difference between this and a person's actual age, called the "retinal age gap", and whether this may be associated with an increased risk of death.

Deep learning is a type of machine learning and artificial intelligence (AI) that mimics the way humans acquire certain kinds of knowledge. But unlike classical machine learning algorithms, which are linear, deep learning algorithms are arranged in a hierarchy of increasing complexity.

The researchers used 80,169 fundus images obtained from 46,969 adults aged 40 to 69, all of whom were part of the UK Biobank, a large population study involving more than half a million middle-aged and older residents of the United Kingdom.

About 19,200 fundus images of the right eyes of 11,052 participants in relatively good health at their initial health check were used to test the accuracy of the deep learning model in predicting retinal age.

This showed a strong link between predicted retinal age and actual age, with an overall accuracy of up to 3.5 years. The retinal age gap was then assessed for the remaining 35,917 participants over an average follow-up period of 11 years.

During this time, 1,871 participants (5%) died: 321 (17%) from cardiovascular disease, 1,018 (54.5%) from cancer, and 532 (28.5%) from other causes, including dementia.

The share of "fast agers" — those whose retinas looked older than their actual age — with a retinal age gap greater than 3, 5 and 10 years was 51%, 28% and 4.5%, respectively. A larger retinal age gap in years was significantly associated with a 49–67% higher risk of death, excluding cardiovascular disease or cancer.

Each additional year of retinal age gap was associated with a 2% increase in the risk of death from any cause and a 3% increase in the risk of death from specific causes other than cardiovascular disease and cancer, after accounting for potentially influential factors such as high blood pressure, weight, lifestyle and ethnicity. The same process applied to the left eye produced similar results.

This is an observational study, so a cause cannot be established. The researchers also acknowledge that the retinal images were taken at a single point in time and that the participants may not reflect the UK population as a whole.

Nevertheless, they reported: "Our new findings established that retinal age gap is an independent predictor of increased mortality risk, especially non-cardiovascular/non-cancer mortality. These results suggest that retinal age may be a clinically meaningful biomarker of aging."

They added: "The retina offers a unique, accessible 'window' to assess the underlying pathological processes of systemic vascular and neurological diseases that are associated with increased mortality risk. This hypothesis is supported by previous studies suggesting that retinal imaging contains information about cardiovascular risk factors, chronic kidney disease and systemic biomarkers."

The new results, combined with previous studies, lend weight to the "hypothesis that the retina plays an important role in the aging process and is sensitive to the cumulative damage of aging that increases mortality risk."