Retina analysis with AI could help detect about 100 diseases

A technician from the company Diagnos takes a photo of a retina during a demonstration aimed at showing how AI could be used as a tool to quickly screen for more than a hundred diseases, at the company’s offices in Brossard, Quebec, on Tuesday, June 9, 2026. (Christinne Muschi/The Canadian Press)

Montreal researchers say they believe that an artificial intelligence (AI) analysis of a photo of the retina could help detect around 100 diseases, sometimes even before the first symptoms appear.

The list isn’t limited to eye diseases, they say, but could also include cardiovascular, systemic and neurodegenerative diseases.

“We want to help (clinicians) with rapid, population-scale screening for eye and systemic diseases to prevent vision loss or other potentially dangerous conditions,” said Professor Ismail Ben Ayed, a researcher with the École de technologie supérieure (ÉTS) who discussed his work exclusively with The Canadian Press on the sidelines of the official launch of the ÉTS–DIAGNOS Chair.

Ayed and his colleagues at Diagnos, where he serves as chief scientist responsible for artificial intelligence, say they believe this approach could significantly improve the speed of screening and help prevent serious complications, particularly vision loss.

To date, the AI has been trained using some 500,000 images provided by patients.

Specifically, he explains, the technology allows for the analysis of retinal images taken in clinical settings using equipment already widely used by optometrists and ophthalmologists.

“These are the same images that they (clinicians) would have. It’s how we process them that’s different,” explained Ayed, whose research areas include medical image analysis and computer-aided diagnosis.

“We take the images captured during their comprehensive eye exam, analyze them, and then provide that information back to them, and the good news is that it contains plenty of biomarkers, and not just for eye health.”

The machine is capable of “extracting very complex information that cannot be seen with the naked eye, such as measurements of vascular structures that appear to be linked to cardiovascular health,” he added.

The goal is not to replace specialists, he assures, but rather offer them a rapid triage tool to identify high-risk cases more quickly and speed up access to diagnosis.

What makes this particular AI unique is that it can “explain” to clinicians how it reached its conclusion.

For example, it can indicate the areas of the image that support its conclusions and assess its own level of uncertainty, leaving the responsibility for final diagnosis to health care staff.

“We want to make (the AI) explainable,” said Diagnos COO Yves-Stéphane Couture, whose firm uses artificial intelligence to provide information to health care professionals. “We want to be able to understand how it made its decision, what it based it on.”

This is a “very important” feature for clinicians who must then come face-to-face with patients, he said.

“The clinician needs to be able to understand what the AI did so they can then explain it to their patient,” Couture notes.

Researchers are also developing mechanisms to allow the tool to tell clinicians its level of confidence in its conclusions.

“The AI will be able to say that it has a high level of confidence, or even say, ‘I have low confidence based on what I’ve seen and how I was calibrated; I’m not at all certain, so be careful,’” said Ayed.

Nevertheless, he maintains that AI has shown great maturity when it comes to screening for eye diseases.

In recent years, the U.S. Food and Drug Administration (FDA) has authorized the use of various AI technologies to screen for diabetic retinopathy.

Significant progress has also been made in screening for conditions like glaucoma and age-related macular degeneration.

It is also believed that AI could contribute to the screening of cardiovascular diseases like hypertension or atherosclerosis, or neurodegenerative disorders like Alzheimer’s, multiple sclerosis or Parkinson’s.

“We need to be clear about what we’re offering because people’s health is at stake,” Couture concluded. “Making mistakes in health care puts patients at risk. I’m very, very cautious about that.”

This report by The Canadian Press was first published in French on June 12, 2026