AI eye scan may help detect dementia and diabetic nerve damage early, says Prof Malik

WASHINGTON: A British-Pakistani scientist based in Qatar says an artificial intelligence-assisted eye scan that takes only a few minutes could help identify dementia and diabetic nerve damage years before symptoms become visible.

Professor Rayaz Malik of Weill Cornell Medicine-Qatar told Dawn that researchers are using corneal confocal microscopy (CCM) to detect nerve damage at an early stage. He said the technology could improve diagnosis and treatment, particularly in countries such as Pakistan where diabetes is rising sharply.

Professor Malik, who works on diabetic neuropathy and neurodegenerative diseases, said the cornea offers an important view into disorders affecting the nervous system because of its dense nerve supply. He explained that CCM was traditionally used by eye specialists to identify infections and abnormalities on the front surface of the eye, but he and his colleagues found that it could also reveal microscopic nerve fibre damage linked to conditions beyond the eye.

Referring to the scientific basis for the approach, he said:

“The cornea has the richest sensory innervation in the body,”

He added:

“We evolved to protect vision, so the cornea contains a dense network of nerves that can reveal damage occurring elsewhere in the body.”

Research development over two decades

Professor Malik said the work began in 2001 after a discussion with Nathan Efron while he was studying diabetic neuropathy, a condition caused by nerve damage in people with diabetes. He recalled that the idea initially faced scepticism.

“Many experts thought we were crazy,”

He said the team later published a landmark study in 2003.

“But in 2003, we published the first paper showing corneal nerve loss in patients with diabetic peripheral neuropathy.”

According to Professor Malik, that study helped launch more than 20 years of further research. Since then, researchers in different countries have shown that CCM can detect nerve damage associated with diabetes, chemotherapy, inflammatory conditions and infectious diseases.

He said research carried out in Qatar over the past 12 years has also shown that the technology can identify neurodegeneration linked to dementia, Parkinson’s disease, multiple sclerosis, schizophrenia and autism.

Potential role in dementia detection

Professor Malik said one of the most important uses of the technology may be in spotting dementia long before memory problems appear. He said that by the time patients are diagnosed after presenting with memory loss, the underlying nerve damage has often been progressing for a decade or more.

“When patients come to the doctor with memory loss and are diagnosed with dementia, the underlying nerve damage has usually been developing for 10 to 15 years,”

He added:

“At that stage, treatments are largely ineffective.”

He said the key challenge is to identify patients during the earliest stage of cognitive decline, known as mild cognitive impairment. According to him, MRI scans generally become abnormal only in advanced dementia, whereas CCM has already shown abnormalities in some people with mild cognitive impairment. He said many such patients go on to develop dementia within five years.

“MRI scans usually become abnormal only in advanced dementia,”

He added:

“But we have shown that some people with mild cognitive impairment already have abnormal corneal nerves on CCM, and many of these patients develop dementia within five years.”

Use in diabetes care and AI support

Professor Malik said CCM can also detect diabetic neuropathy up to five years earlier. He said this matters because early intervention can help repair nerves through weight loss, blood glucose control, lower lipid levels and reduced blood pressure.

“This is crucial because we now know that weight loss, controlling blood glucose, lowering lipids and reducing blood pressure can actually lead to nerve repair — especially when treatment begins early,”

He said artificial intelligence has significantly increased the usefulness of the technology by reducing analysis time from lengthy manual review to processing in seconds.

“AI has been a game changer,”

Professor Malik said:

“When I look at a nerve image, I may identify three or four features. AI can analyse more than 2,500 features and learn patterns associated with specific diseases.”

According to him, AI systems can now identify the underlying neurodegenerative disease with 90 to 95 per cent certainty. He also said that in some studies on diabetic neuropathy and Parkinson’s disease, researchers achieved nearly 100pc sensitivity and specificity.

Barriers to wider adoption

Professor Malik said one obstacle to broader use has been resistance from some specialists.

“One major roadblock has been convincing neurologists that an endocrinologist could use an eye scan to diagnose neurological disease,”

He added:

“Twenty-five years later, we have largely convinced them, although there are still some dinosaurs who refuse to believe — and we all know what happened to the dinosaurs.”

He also said the limited availability of CCM machines has slowed adoption, noting that for years only one German manufacturer made the equipment. However, he said a Chinese company has now started producing new CCM devices, which could improve access and lower costs globally. He said this could be especially relevant for developing countries such as Pakistan, where diabetes is increasing and health systems face pressure from chronic disease.

Professor Malik also noted that the histories of artificial intelligence and confocal microscopy intersect in the work of Marvin Minsky of the Massachusetts Institute of Technology, who was a pioneer of AI and also invented the confocal microscope in the 1950s.