Researchers at King’s College London have developed a powerful new AI model that can predict a person’s risk of vision-threatening diabetic retinopathy (DR) up to three years in advance.
Using over 1.2 million retinal images from NHS patients with diabetes, the model is designed to help doctors identify who needs urgent monitoring and who might safely skip some screenings—potentially saving sight, time, and money.
The study, recently published in the journal Communications Medicine, marks a major step forward in the use of artificial intelligence in healthcare. It also offers a new way to improve the efficiency of the NHS Diabetic Eye Screening Program (DESP), which currently screens more than 3.2 million people every year in England, at a cost of over £85 million.
What is Diabetic Retinopathy?
Diabetic retinopathy is a complication of diabetes that damages the retina—the light-sensitive tissue at the back of the eye. It affects about 1 in 3 people with diabetes and is one of the leading causes of vision loss in working-age adults.
Under the current system in the UK, everyone aged 12 and over with diabetes is invited for an annual eye check, regardless of their personal level of risk. While this has helped prevent many cases of blindness, it also means that millions of people at low risk are screened every year, which can place a heavy burden on both the healthcare system and patients.
How the AI Model Works
The King’s College team trained their AI using anonymized data from over 100,000 people with diabetes in South East London. This included 1.2 million eye images taken during routine diabetic eye screenings.
They then tested the model using a separate dataset of about 70,000 images from the INSIGHT Health Data Research Hub, to make sure it worked well on a wide range of people and images.
The AI looks for subtle signs in retinal images that might go unnoticed during a normal eye exam but that suggest a person is at risk of developing sight-threatening diabetic retinopathy in one, two, or three years. These kinds of early predictions are not possible with current NHS screening tools.
What This Could Mean for the NHS and Patients
If adopted into practice, this AI tool could allow the NHS to personalize screening schedules. For example:
- People at low risk of vision loss might be safely screened less often (e.g., every 2–3 years).
- Those at high risk could be seen more urgently or referred for early treatment.
This targeted approach could help:
- Reduce unnecessary appointments, saving time for both patients and clinicians.
- Lower NHS costs by reducing over-screening.
- Prevent vision loss by catching high-risk cases earlier.
Professor Timothy Jackson, one of the study’s senior authors, said the project shows the real-world value of using routine NHS health data for AI development. “It also shows the way ahead, in terms of how AI might move from hype to tangible patient benefit,” he said.
Professor Christos Bergeles added that AI has the potential to modernize diabetic eye screening programs without compromising their ability to prevent blindness. “This has tremendous social and economic prospects,” he said.
Review and Analysis
This study is a promising example of how AI can support—not replace—doctors, by helping them deliver more personalized, efficient care. The model doesn’t diagnose disease on its own, but it can flag patients who may need closer attention, giving healthcare providers more time to focus on those at highest risk.
By using data that is already collected as part of routine care, the researchers avoided extra costs or burdens on patients. They also took care to validate their model on a separate dataset, improving its reliability.
However, more real-world testing is still needed before the tool can be rolled out across the NHS. Clinical trials and ethical reviews will be essential to make sure the system is safe, fair, and works for everyone—especially in a diverse population with varying risk factors.
If successful, this kind of AI-driven screening model could set a new standard for how chronic diseases like diabetes are managed—not only in the UK, but around the world.
If you care about blood sugar, please read studies about why blood sugar is high in the morning, and how to cook sweet potatoes without increasing blood sugar.
For more information about brain health, please see recent studies about 9 unhealthy habits that damage your brain, and results showing this stuff in cannabis may protect aging brain, treat Alzheimer’s.
The research findings can be found in Communications Medicine.
Copyright © 2025 Knowridge Science Report. All rights reserved.
