Eye scan can reveal risk of heart attack or stroke

A simple digital photograph of the back of the eye can predict a major cardiovascular event - such as a heart attack or stroke - that will happen within the next decade with 70% accuracy, according to research backed by the British Heart Foundation and the National Institute for Health and Care Research (NIHR).

The researchers believe routine retinal scans could also be used to track a person's heart health over time, as they also found a link between the change in risk score over three years and the likelihood of a major cardiovascular event.

The eye scan is analyzed using artificial intelligence (AI), which provides a personalized risk prognosis in a fraction of a second.

Those at highest risk could be referred to a GP, who might prescribe blood pressure medication or statins to lower cholesterol. In the future, the researchers hope that anyone having an eye exam could receive a notification about their heart condition on their smartphone.

Dr Ify Mordi, a British Heart Foundation research fellow at the University of Dundee and consultant cardiologist, led the study, published in the journal Cardiovascular Diabetology.

He said:

"It may come as a surprise, but the eyes are the window to the heart.
If there is damage or narrowing of the blood vessels at the back of the eye, there is a good chance that this will also be seen in the vessels deeper in the body that supply the heart, which can lead to a heart attack or stroke.
This is a quick scan that is routinely performed and takes less than a minute. It could be an important part of a care package, along with blood pressure and cholesterol checks, to identify people who might benefit from treatment or lifestyle changes."

How the technology works

Researchers at the University of Dundee have developed AI technology to analyse digital retinal photographs that are usually part of routine eye tests.

First, the AI was trained to spot warning signs such as narrowing of blood vessels, blockages, and lesions that could be signs of impending heart problems. Then, a “black box” approach was used, allowing the technology to use deep machine learning to look for any detail in the images, from the size to the location of blood vessels.

After training on ~4,200 images, the black-box AI was tested to see how well it could predict people who would have a heart attack or stroke, or die from cardiovascular disease, over the next decade. The team found that the AI predicted 70% of these cases when it was tested on eye scans from more than 1,200 people.

Some participants also had a repeat scan three years after the first. When the researchers analyzed how the risk scores identified by the AI changed between scans, they found that the fifth of the group with the biggest increase in score had a 54% higher risk of a major cardiovascular event than the rest.

This 54% increase in risk occurred in people whose AI score increased by just 3% over three years – for example, if the 10-year risk of a cardiovascular event increased from 20% to 23%.

Comparison with current methods

The team also compared the AI technology with the risk prediction people currently get from routine health checks with their primary care physician — their “cardiovascular risk,” which is the percentage chance of having a major heart event over the next decade based on factors such as age, gender, blood pressure, cholesterol levels, and smoking. The researchers found that the AI and the traditional method identified nearly the same proportion of people at risk of having a heart attack, stroke, or dying from cardiovascular disease.

When clinical risk, retinal scanning and an additional genetic test were combined, the predictive accuracy increased to 73%. This means that together they could potentially identify an additional three people in every 100 who were at risk.

The AI has been tested on people with diabetes because they have routine retinal scans on the NHS to check for complications affecting the eye. But the researchers believe that analysing the blood vessels in the eye to assess cardiovascular risk should work for most people, not just diabetics.

Conclusion

The study was a collaboration between clinical researchers including Dr Mordi and Dr Alex Doney from the University of Dundee, and the computer scientists who developed the AI, led by Professor Emanuele Trucco and Dr Mohammad Sayed.

Professor Brian Williams, chief scientific and medical officer at the British Heart Foundation, said:

“The more accurately we can predict a person’s risk of having a heart attack or stroke, the greater the chance of preventing them.
Cutting-edge innovations such as using retinal scans alongside medical screening could play a role in improving risk prediction, which is important to achieving the British Heart Foundation’s target of preventing 125,000 heart attacks and strokes in the UK by 2035.
However, more research is needed to confirm the reliability of this predictive accuracy and to determine how realistic it is to incorporate retinal scans into clinical practice.”