Artificial Intelligence in Diabetic Retinopathy

Original Article:

People with diabetes are at significant risk of eye disease

Diabetic retinopathy is a common complication in people with diabetes. There are different degrees of retinopathy, which if left untreated, can progress leading to loss of vision. Fortunately, before visual loss occurs, diabetic retinopathy can be recognized at an early stage by screening. At the  early stage, treatments to prevent visual loss are very effective. About 35% of people with diabetes develop diabetic retinopathy. The challenge is how to ensure that everybody has regular retinal screening – for most people this means every year.

Artificial intelligence (AI) and machine learning can be used to diagnose diabetic retinopathy

Early screening is important for the prevention, diagnosis and treatment of diabetic retinopathy.With the increased use of AI, so-called smart imaging can be used to take a picture of a retina with a computer programmed to detect if retinopathy is present and the severity, screening using AI had a strong agreement with an in-person diagnosis and almost perfectly agreed with tele-ophthalmology screening results.

In this study, participants with diabetes underwent both an in-person and Tele evaluation of retinal images. Retinal images were taken without the need to dilate the pupils. The results were compared with an in-person examination with the retina examined through dilated pupils. 

The researchers found that there was a high level of agreement between both methods including for more severe retinopathy.

What do these findings mean?

Incorporating AI into tele-medicine could greatly increase the number of people who can be screened for diabetic retinopathy, particularly for people who live in rural regions or are far away from an ophthalmologist. This is especially important for identifying disease early and allowing for continued routine screening even at home during the pandemic.