Artificial intelligence to transform diabetic eye testing

People with diabetes will soon have the opportunity to undergo an in-depth eye examination that is powered by artificial intelligence.

Scientists from the University of Liverpool and Manchester Metropolitan University have developed the test to detect hidden eye complications that are linked to diabetes, such as diabetic peripheral neuropathy.

By altering equipment that is traditionally used by optometrists on the high street, the researchers are hoping this test will become a common screening tool for diabetes-related eye problems.

The artificial intelligence (AI) element of the test has the ability to predict future damage by scanning the front of the eye rather than the back.

Lead author Dr Uazman Alam said: “What we know from a body of work which I’ve been very heavily involved in over the past 15 to 20 years, is that the nerves at the front of the eye actually reflect nerve damage elsewhere in the body.”

Diabetic peripheral neuropathy is the biggest cause of limb amputation in people living with diabetes.

The condition is triggered when high blood sugar levels damage the nerves that send messages from the brain and spinal cord to the rest of the body.

Researchers have now received a £1.4 million funding grant to create a new machine by essentially redesigning the optical coherence tomography (OCT) device – a tool commonly used by optometrists.

Dr Alam noted: “Our device will scan the front of the eyes. The nerves are very small and therefore it needs a higher resolution. It will also have AI embedded in it and could save clinicians time and save the NHS money.”

The 10-gram monofilament test is currently used to identify sensory loss in the limbs of people living with diabetes; however, it can miss a lot of people.

“At the moment, [patients] are screened, but the tests we use aren’t sensitive. This we’re hoping will be a lot more sensitive,” said Dr Alam.

He added: “Rather than having to take measurements of the nerves, we can use the entire image to detect the nerve damage and actually predict those who will develop it.

“In nearly all countries, diabetes is going to increase. For us to have somebody individually undertake the [10 gram monofilament] test and detect nerve damage is quite time consuming and would be a major economic burden to healthcare services.”

Researchers are hoping that the study will end in 2027, ultimately resulting in a pilot clinical validation trial in healthy volunteers with diabetes at Aintree University Hospital in Liverpool.

Dr Alam said: “AI will be an important facet of all healthcare systems at some point but will need developing further before it is adopted widely.

“I think we have to remember that AI is not just the images that we’re talking about, but it can also be data as well.”

He concluded: “It’s here to stay. We need to develop it in a way that is ethical. I think it’s important and I think it probably actually needs to be taught in medical schools as well. It’s going to be entrenched within the healthcare system.”