AI-powered eye study uncovers key links between retina and diseases

AUSTRALIA – A groundbreaking international study led by the Walter and Eliza Hall Institute of Medical Research (WEHI) in Australia has revealed new insights into how the retina is connected to various diseases.

By using artificial intelligence (AI) to analyze eye images from over 50,000 people, researchers have created the most detailed map of retinal thickness ever recorded.

How AI helped unlock retinal secrets

The study, funded by the Lowy Medical Research Institute in California, used advanced AI technology known as a convolutional neural network to analyze optical coherence tomography (OCT) scans from approximately 54,000 individuals stored in the UK Biobank.

This AI system generated high-resolution retinal maps, measuring over 29,000 specific points across the retina, giving researchers an unprecedented view of its structure.

The study was a global collaboration, with researchers from Moorfields Eye Hospital and University College London (UK) and the University of Washington (USA) contributing to the findings.

Key findings: Retina’s role in disease detection

Earlier research has shown that retinal thickness is linked to diseases, but this AI-driven study provided much deeper insights into these relationships.

One of the most significant discoveries, published in Nature Communications, is that reduced retinal thickness is strongly linked to multiple sclerosis (MS), a serious neurological condition affecting the brain and spine.

The study confirmed previous reports that OCT imaging could be a reliable biomarker for detecting and tracking MS progression.

Moreover, retinal thinning was also associated with other neurodegenerative conditions and metabolic disorders, including cardiovascular diseases and diabetes.

The study noted that the retina has “unique metabolic sensitivities,” meaning that changes in its thickness may indicate disruptions in the body’s metabolic processes.

Researchers also uncovered genetic factors influencing retinal thickness, identifying at least 294 genes involved in the retina’s growth and development.

This discovery provides further clues into how genetics play a role in eye health and systemic diseases.

Why these findings matter

This research highlights the potential of routine eye scans as a simple, non-invasive way to detect and monitor diseases affecting the brain and body.

Since the retina is an extension of the central nervous system, eye imaging could become a powerful tool for identifying conditions like dementia, multiple sclerosis, and diabetes early.

Lead researcher Dr. Vicki Jackson from WEHI emphasized this point, stating: “We’ve shown that retinal imaging can act as a window to the brain, by detecting associations with neurological disorders like multiple sclerosis and many other conditions.”

The study also reinforces the growing field of oculomics, which focuses on using eye imaging to predict and diagnose diseases non-invasively.

By identifying specific areas of the retina that change in response to different diseases, researchers believe this technique could revolutionize early detection and treatment strategies.

The bigger picture: AI in eye health

AI-driven eye studies are becoming a global trend, with various institutions developing new technologies to assess and predict health conditions.

• In Australia, another study found that damage to the conjunctiva (the eye’s outer layer) could predict skin cancer, leading to the development of an AI-powered system that detects sun-related eye damage using mobile and desktop applications.
• In Singapore, the Singapore Eye Research Institute has developed AI solutions to screen for chronic kidney disease and predict biological age using retinal scans.
• In China, researchers created VisionFM, a generative AI model trained on 3.4 million eye images from over 500,000 people worldwide. This model is capable of automated eye disease diagnosis, tracking disease progression, predicting intracranial tumors, and analyzing systemic biomarkers through eye imaging.