According to a story from Ophthalmology Times, the COVID-19 pandemic rapidly expanded the usage of telemedicine, including virtual, tele-ophthalmology appointments. It also highlighted the effectiveness of AI, and both of these methods could play a role in the continuing treatment and diagnosis of diabetic retinopathy. A recent analysis sought to compare the efficiency of AI versus a retina specialist working virtually to diagnose diabetic retinopathy.

About Diabetic Retinopathy

Diabetic retinopathy, sometimes called diabetic eye disease, is a condition in which the retina sustains damage as a result of diabetes. Nearly 80 percent of patients who live with either type of diabetes are affected after 20 years. Prompt treatment can greatly reduce the risk of progression, but the condition remains a major cause of vision loss in developed countries. In the US, it accounts for 12 percent of all new blindness cases and is the leading cause for people age 20-64. The condition may not inflict any symptoms at first, but most people eventually have blurry vision, vision loss, distorted or darkened images, and eventually blindness. While there is no cure for the disease, it can be treated very effectively; methods include injected corticosteroids/anti-VEGF agents, vitrectomy, and laser surgery. These treatments, when done early in the disease course, can prevent vision loss. To learn more about diabetic retinopathy, click here. 

About The Study

The analysis included 40 patients for a total of 80 eyes. First the eyes were assessed for the presence of diabetic retinopathy by AI. Then they were assessed via tele-ophthalmology and in-person by retina specialists. 26 eyes were un-gradable by AI and 11 were un-gradable by tele-ophthalmology. Nevertheless, of the eyes that were graded, AI showed close agreement with in-person inspection and nearly identical agreement with tele-ophthalmology. As a result, all three methods were a close match in effectiveness. 

The researchers concluded that tele-ophthalmology can be a reliable tool for diagnosis of diabetic retinopathy; however, they also noted the weaknesses of both this approach and the AI, as a significant number of the eyes could not be graded by these methods. The team concludes that further research should strive to reduce the number of un-gradable eyes.