Eye for Diabetes asks citizens to help train new software
The citizen science project Eye for Diabetes needs 5,000 people to record their retinal scans in a database so a computer programme can learn to predict eye diseases
Lend an eye
Diabetic retinopathy is a common complication of diabetes, which occurs when a high blood sugar level causes damage to the walls of retinal blood vessels over time. This goes unnoticed in the beginning, but the damage can eventually lead to partial and even total blindness.
Diabetic retinopathy affects about one-third of people with diabetes, and the risk of retinal damage increases the longer the disease goes undetected. “There are treatments to stop the condition from worsening, such as laser therapy, but the loss of sight already experienced cannot be repaired,” explains Patrick De Boever of Vito, Flanders’ institute devoted to technological research. “So it’s crucial to spot the disease at an early stage.”
Say cheese
That’s why the Mona research team at Vito is developing software to automatically detect symptoms of the disease. Mona software will facilitate and upgrade analyses by ophthalmologists but can also simplify screenings at opticians and first-line health-care providers, like general practitioners and pharmacists.
A digital picture of the back of the eye, taken with a retinal camera, could be checked in just a few minutes. To optimise the software’s detection skills, Vito set up the citizen science project Eye for Diabetes in collaboration with research group Smit (linked to nanotech research centre imec and the VUB), Brussels science communication centre Wtnschp and the Diabetes League association.
Eye for Diabetes was selected by the government of Flanders for support as part of its citizen science projects. There is an online platform (in English and Dutch) where everyone can easily participate in the research by letting a medical professional input data from their retinal scan.
Our computer programme learns to recognise the eye problems in the same way a child learns, through repetition
When a retinal image of a patient’s eye is taken, professionals are asked to record any lesions such as haemorrhages – visible as red spots. They are an indication of diabetic retinopathy.
“Our computer programme learns to recognise the eye problems in the same way a child learns, through repetition,” explains De Boever, who is co-ordinating Eye for Diabetes. “We have a large database of retinal images to enhance the intelligence of our software, and we hope that citizens can help us to process some 5,000 images, which would already provide a strong database.”
The platform is user-friendly, but the project team is looking to further boost the user experience and interest by introducing game features. The team is also presenting the initiative at several events, such as during Flanders’ Dag van de Zorg, or Care Day, on 17 March. The researchers will provide info and demonstrations at the Jan Portaels Hospital in Vilvoorde.
Bright future
Apart from involving citizens in the research, the project also aims to raise more awareness about diabetes in general, which currently affects about one in 12 Belgians. There is a strong link between type 2 diabetes, the most common form of the disease, and an unhealthy lifestyle. Risk factors include a high-calorie diet, insufficient exercise and smoking.
The Mona team is also developing similar computer models for other eye diseases, like glaucoma. But De Boever sees potential for future software that could detect other health problems such as high blood pressure, the risk of stroke and Alzheimer’s disease.
“Because our eyes are directly linked to our brain, the eye forms a kind of window to the brain and, indirectly, also to the heart,” he says. “This allows us to see signs of many possible health issues in the entire body.”
The Eye for Diabetes project runs until March of 2020.
Photos: The yellow spots on this retina scan shows that the patient is suffering from diabetic retinopathy (top), people could get their retinas scanned during the Supernova tech event last autumn (above)
©Courtesy Vito
