AI: Can it help prevent blindness in diabetics?

This is the second feature in a six-part series looking at how AI is changing medical research and treatment.

Terry Quinn was just a teenager when he was diagnosed with diabetes. In some ways he rebelled against the label and repeated tests, not wanting to feel different.

His greatest fear was that one day his foot would have to be amputated. Vision loss, another potential complication of diabetes, wasn’t really on her radar. “I never thought I’d lose my sight,” says Quinn, who lives in West Yorkshire.

But one day he noticed that his eye was bleeding. Doctors told him he had diabetic retinopathy: damage to blood vessels in the retina related to diabetes. This required laser treatment and then injections.

Eventually the treatments were not enough to stop his vision from deteriorating. Walking into lamp posts will hurt his shoulder. He couldn’t make out his son’s face. And he had to give up driving.

“I felt pathetic. I felt like a shadow of a man who couldn’t do anything,” he recalls.

One thing that helped her overcome her despair was the support of the Guide Dogs for the Blind Association, which paired her with a black Labrador named Spencer. “It saved my life,” says Quinn, now a fundraiser for guide dogs.

In Great Britain NHS invites patients. For a diabetic eye screening every one or two years.

U.S. guidelines state that every adult with type 2 diabetes should be screened at the time of diabetes diagnosis, and then annually if there are no problems. Yet for many, in practice this is not the case.

“There is clear evidence that screening prevents vision loss,” says Romasa Chana, a retina specialist at the University of Wisconsin-Madison in the US.

US barriers include cost, communication and convenience. Dr Chana believes that making the tests easier to access will help patients.

To screen for diabetic retinopathy, healthcare professionals take pictures of the back inner wall of the eye, called the fundus.

Currently, manually interpreting fundus images is “too repetitive,” says Dr. Chana.

But some believe artificial intelligence (AI) could speed up the process and make it cheaper.

Diabetic retinopathy develops in fairly distinct stages, which means AI can be trained to pick it up.

In some cases, AI can decide whether to refer to an ophthalmologist, or work with human image graders.

One such system has been developed by Retmarker, a Portugal-based health technology company.

Its system identifies fundus images that may be problematic and sends them to a human expert for further investigation.

“Usually we use it as a support tool to give information to a human to make a decision,” says João Diogo Ramos, chief executive of Retmarker.

He believes that fear of change is limiting the use of such AI-powered diagnostic tools.

Independent studies have suggested that systems such as Retmarker screening and Eyenuk’s EyeArt have acceptable rates of sensitivity and specificity.

Sensitivity is how good a test is at detecting disease, while specificity is how good it is at detecting the absence of disease.

In general, higher sensitivity can be correlated with more false positives. False positives cause both anxiety and expense, as they lead to unnecessary specialist visits. In general, poor quality images can lead to false positives in AI systems.

Google Health researchers are examining the weaknesses of an AI system they developed to detect diabetic retinopathy.

It performed very differently when trialled in Thailand compared to the hypothetical scenarios.

One problem is that the algorithm requires primitive fundus images. It was a far cry from the realities of occasional dirty lenses, unpredictable lighting, and varying levels of training for camera operators.

The researchers say they learned lessons about the importance of working with better data and consulting a wider range of people.

Google has enough confidence in its model that in October the company announced it was licensing it to partners in Thailand and India. Google also said it is working with the Thai Ministry of Public Health to evaluate the cost-effectiveness of the tool.

Cost is a very important aspect of new technology.

Mr Ramos says Retmarker’s service can cost around €5 per screening, although with variations depending on volume and location. In the US, medical billing codes are quite extensive.

In Singapore, Daniel SW Ting and colleagues compared the costs of three models of diabetic retinopathy screening.

The most expensive was human evaluation. However, full automation was not the cheapest, as it had more false positives.

The most affordable was a hybrid model, where initial filtering of results was done by AI before humans took over.

The model has now been integrated into the Singapore Health Service’s national IT platform and will go live in 2025.

However, Professor Ting believes that Singapore was able to achieve cost savings because it already had a strong infrastructure for diabetic retinopathy screening.

The cost-effectiveness is therefore likely to vary greatly.

Bilal Mateen, chief AI officer at the health NGO PATH, says cost-effectiveness data around AI tools to protect sight is insufficient in rich countries like the UK or a few middle-income countries like China. is strong But not so for the rest of the world.

“With rapid advances in what AI is capable of, we need to ask less if it’s possible, and more if we’re building for all or just the privileged few. For effective decision-making, we only need data on effectiveness,” emphasized Dr. Mateen.

Dr. Chana also points to a health equity gap within the U.S., which she hopes the tech can help bridge. “We need to expand this to places with limited access to eye care.”

She also emphasizes that older people and people with vision problems should see eye doctors, and that AI’s ability to routinely detect diabetic eye disease should not address all other eye diseases. It should be stopped. Other eye conditions, such as myopia and glaucoma, have proven more difficult for AI algorithms to detect.

But despite these caveats, “the technology is very exciting,” says Dr. Chana.

“I would love to see timely screening of all my diabetes patients. And I think that, given the burden of diabetes, it’s a really potentially great solution.”

Back in Yorkshire, Mr Quinn certainly hopes the new take will take off.

If AI had been there to detect his diabetic retinopathy earlier, “I would have grabbed it with both hands.”