Analysis of the ‘fingerprint’ of blood vessels in the retina could make it possible for people to keep tabs on their cardiovascular health during routine eye tests.
Summary
An artificial-intelligence (AI) algorithm analyses the unique pattern of tiny blood vessels at the back of the eye to predict if someone is likely to have a stroke in the future. The computer program was developed with eye scans from nearly 69,000 UK Biobank participants. Similar programs could eventually make it possible to combine general health checkups with routine eye tests.
An artificial-intelligence (AI) algorithm can predict whether someone is likely to have a stroke in the future by analysing the unique pattern of tiny blood vessels at the back of the eye. The computer program, developed with eye scans from nearly 69,000 UK Biobank participants, works as well as the blood tests people do as part of health checks.
I hate to use the cliché, but really, [the eye] is a window into general health – it goes far beyond vision.
Dr Siegfried Wagner, University College London, UK
Similar algorithms could analyse images taken during routine eye tests, which might encourage more people to regularly check on their cardiovascular health – though questions remain about how this could be implemented with private optometrists.
The tiny blood vessels in the retina, the light-sensitive layer in the back of the eye, reflect the condition of the wider cardiovascular system. “I hate to use the cliché, but really, [the eye] is a window into general health – it goes far beyond vision,” says Siegfried Wagner from University College London, UK, who wasn’t involved in the new study. His team and others have previously analysed UK Biobank retinal images to detect high blood pressure, diabetes, heart failure and more.
Eye-opening stroke prediction
A team of researchers have now tasked an AI program with looking at retinal images from almost 69,000 UK Biobank participants – including some who had a stroke in the years after the image was taken.
It’s another really good example of how ophthalmic eye imaging can be a real powerful vehicle for understanding systemic health and disease.
Dr Siegfried Wagner, University College London, UK
They found notable differences in the ‘fingerprint’ of blood vessels between people who had a stroke and those who didn’t. “It’s another really good example of how ophthalmic eye imaging can be a real powerful vehicle for understanding systemic health and disease,” Wagner says.
When taking into account someone’s sex and age, retinal analysis was as good at predicting strokes as methods that require blood tests for glucose and cholesterol levels.
Although the program is not ready for the clinic yet – it needs to be tested with people beyond UK Biobank – “it has huge potential”, says study co-leader Mayinuer Yusufu from the University of Melbourne, Australia. Retinal images are cheap and non-invasive, she says, and could be done during routine eye tests. And eventually they could even be used to identify other common conditions, for example chronic kidney disease.
An eye test for heart health
Wagner adds that even if retinal analysis won’t ever be as accurate as a full health check that includes blood tests, “it reaches a much wider audience – people who don’t realise their cardiovascular risk is really high”. Just one in 12 people between 40 and 74 currently attend their regular NHS health checks while nearly 80% of the population regularly sees an optometrist for eye tests.
“It’s a great proof of concept but actually implementing that into patient impact is still a while away,” Wagner warns. There are practical issues, such as connecting retinal-imaging machines to a network that can run an AI program. And in Wagner’s experience, optometrists are worried about being responsible for health analyses when it is not part of their usual duties of care.
Some algorithms that analyse health through retinal images are already available to buy. “The question now is not: ‘Can it be done?’” Wagner says. “It’s more about ‘How do we implement it?’”
