Detecting early signs of Diabetic Retinopathy (DR) in retina images

Diabetes

Diabetes is a disease that occurs when your blood glucose, also called blood sugar, is too high. Blood glucose is your main source of energy and comes from the food you eat. Insulin, a hormone made by the pancreas, helps glucose from food get into your cells to be used for energy. Sometimes your body doesn’t make enough —or any— insulin or doesn’t use insulin well. Glucose then stays in your blood and doesn’t reach your cells. Over time, having too much glucose in your blood can cause health problems. Although diabetes has no cure, you can take steps to manage your sugar levels and stay healthy.

Diabetic retinopathy

Diabetic retinopathy is a diabetes complication that affects the eyes. High sugar levels in the blood can lead towards the blocking of the tiny blood vessels in the retina. The retina is the light-sensitive layer of the eye on which the sight is projected and then transmitted to the brain. The damage to small vessels in the retina is called retinopathy. In the first stage, small widening and weak spots in the blood vessels occur, causing minor bleeding and fluid leakage.

Most people experience no discomfort at first, but urgent treatment is needed at this stage. Hence the necessary (annual) examination. The eye doctor is able to determine the injuries, even before you are experiencing discomfort. Later symptoms of diabetic retinopathy may include blurred or fluctuating vision, impaired color vision, spots or dark strings in your vision, and vision loss. Anyone who has diabetes can develop diabetic retinopathy.

If diabetes is not controlled in a proper way, the small capillaries can clog up in a next stage of the disease. This results in oxygen deficiency in various places in the retina. Subsequently, new capillaries are created that are of very poor quality. They burst easily and can cause pronounced bleeding. A patient can suddenly notice a dark spot, or can lose sight for months due to the thick blood curtain. If nothing is done at this stage, the newly formed blood vessels will shrivel up and eventually tear off the retina, resulting in definitive blindness. Thanks to a strict sugar and blood pressure check and annual preventive examinations at the eye doctor, the risk can be limited to the strict minimum.

Goal - Why we need your help!

The "Eye for diabetes" ("Oog voor diabetes" in Dutch) project has the goal to create a computer model based on artificial intelligence to be able to detect early signs of diabetic retinopathy. This computer model can only be trained successfully by supplying it with a huge amount of example images on which the signs of diabetic retinopathy have been labeled by humans. Early signs of diabetic retinopathy can be subtle and difficult to spot, even for well-trained specialists. We need your help in our search to find specific signs of diabetic retinopathy in a public dataset from EyePacs (a screening network in the US).

Your help will allow us to make a significant difference to the detection and follow up of diabetic retinopathy, and will better support the eye health of all patients suffering from diabetes.

Approach

Thanks to the progress recently made in the domain of artificial intelligence, computers can detect diseases in medical images with the accuracy of a human specialist. Training artificial intelligence with this kind of quality requires a large amount of high quality examples. A few public datasets showing signs of diabetic retinopathy have already been created by ophthalmologists but they are way too small to obtain good results. We need at least 100 times the amount of labeled images. This kind of work can not be done by a few specialists, but it can be done by a large group of people!

The EyePACS screening network has made a large dataset of retinal images from diabetes patients publicly available. EyePACS has given us permission to use their dataset to label the lesions, caused by diabetes, on them. To the best of our knowledge, the EyePACS dataset is the largest, public available, DR screening dataset from a diverse number of screening centers. It has been used in numerous scientific publications, and is our data set of choice to create better labels for.

All labels of signs of diabetic retinopathy created in this project will be made publicly available. This will allow ALL patients with diabetes, no matter in which country, to benefit from our work.

Funding support

This project is made possible through the support of the Department Economy, Science and Innovation of the Flemish Government (Belgium), and granted funding in the first citizen science call of 2018.