Changing how we see diabetic retinopathy

  • Posted on 15.11.2016

Changing how we see diabetic retinopathy


Edouard Colas

DreamUp Vision Chief Medical Officer


Consider this: about 60 million people in Europe have diabetes. Out of those who have it for more than 20 years, 75% will develop some form of diabetic retinopathy (DR).

It’s a startling statistic. Early detection of DR can prevent blindness, which is why people with diabetes should be tested every year.

But because so many people have diabetes – and the numbers are rising – screening everyone for this debilitating eye disease is a huge challenge. For one thing, we do not have enough ophthalmologists to check the millions of eyes at risk of DR. And even if we did, the cost would be significant.

New approaches are urgently needed. Fortunately, there is a solution on the horizon. By combining sophisticated cameras with artificial intelligence (AI), we can make diabetic retinopathy screening more efficient and cheaper – helping to deal with the growing demand for this crucial service.

‘Deep learning’ is a powerful kind of AI that can detect specific features in an image of the eye with high sensitivity. It allows health professionals to diagnose the stages of retinopathy in milliseconds.

At DreamUp Vision, we are using this technology as a SaaS platform, as well as integrating it into a fundus camera – the kind of camera that ophthalmologists use to scan the eye.

The technology is so flexible that any healthcare professional could scan a patient’s eye and get an immediate answer if the patient has signs of the disease or not. This could go a long way to addressing the shortage of ophthalmologists, while bringing expert care to people who do not live near specialist health centres.

Learn by doing

The incredible thing about this kind of AI is that it learns: the more eye scans it sees, the more accurate it becomes. We are currently collecting data to fine-tune the algorithm for DR diagnosis. After that, we hope to apply the same principle to other eye diseases.

Once the system has ‘learned’ how to recognise DR, it can be retrained to detect glaucoma, cataracts and age-related macular degeneration – all major challenges for an ageing population.

One of the most powerful impacts of AI in healthcare is its power to bring increasing sophistication without driving up costs. Indeed, the biggest expense is the hardware (the fundus camera) while the software is relatively inexpensive.

Even hardware is becoming cheaper and more mobile every day. Soon you will be able to buy an add-on that plugs into your smartphone. Think of the impact this could have if community health workers had access to a mobile camera and deep learning software.

We are looking at a real revolution here with the power to radically decentralise healthcare and address inequality. And, far from replacing health professionals, we expect this to help doctors, nurses and others to be far more efficient in their work.

Patient experience

For patients, their experience will be transformed. Imagine being a patient with diabetes, and perhaps other medical conditions, and having to travel for hours for an annual eye exam by a busy specialist. AI technologies have the power to bring the check-up to you.

For the health system – and the wider economy – swift and early diagnosis can keep people healthy and active for longer by reducing their risk of blindness. It’s a win-win.

We believe artificial intelligence in healthcare is here to help doctors and patients: by automating the screening process, we will reduce the delay to see a specialist, and improve the outcome.

Of course, there is work to be done to make this a reality. We look forward to discussing AI technology and its implications at the MedTech Forum, taking place on 1st and 2nd of December in Brussels.

This blog is part of the MedTech Forum blog series. DreamUp Vision were present on the stage of MedTech Forum 2016. You can follow the conversation under #MTF2016 and find more details and at

The comments are closed.