AI system may allow accurate retinopathy diagnosis by non-ophthalmologists

New research demonstrates that an automated, artificial intelligence (AI) screening system accurately detects diabetic retinopathy in more than 95% of cases. The system can be used by non-ophthalmologists and provides readings in 60 seconds.

Accurate, automated retinal screening is an important development for the millions of diabetic patients who require annual examination for sight-threatening diabetic retinopathy. New research, presented at the 123^rd Annual Meeting of the American Academy of Ophthalmology (12—15 October 2019, CA, USA) — demonstrates that an automated, artificial intelligence (AI) screening system can detect diabetic retinopathy with more than 95% accuracy.

Approximately 25% of diabetic individuals in the USA will develop retinopathy. Initially, such retinal damage may be asymptomatic, but can ultimately lead to blindness.

Frequently asked questions:

• What is diabetic retinopathy?
• How is retinopathy diagnosed?

There are various options available for the treatment of diabetic retinopathy, however, early diagnosis significantly improves treatment effectiveness. As a result, ophthalmologists recommend annual retinal screening for diabetic patients.

EyeArt® is an automated, AI technology for detecting diabetic retinopathy. Such a system may greatly benefit clinicians by enabling accurate, rapid disease identification amongst an ever-increasing diabetic patient population.

To assess the usefulness of the system, researchers directly compared the diagnostic accuracy of EyeArt against the Early Treatment Diabetic Retinopathy Study test — considered the gold standard method for diagnosing diabetic retinopathy.

When tested on 893 diabetic patients, across 15 healthcare practices, the sensitivity of EyeArt was determined to be 95.5% for non-dilated pupil images. This sensitivity increased to 97.4% when the system was challenged with dilated pupil images.

More than 90% of the images that EyeArt deemed to display diabetic retinopathy were confirmed to portray retinopathy, or another ophthalmic condition, by certified graders.

Lead study author Srinivas Sadda, from University of California, Los Angeles’ Doheny Eye Institute (CA, USA), commented: “Diabetic patients already outnumber practicing ophthalmologists in the United States, and unfortunately, that imbalance is only expected to grow. Accurate, real-time diagnosis holds great promise for the millions of patients living with diabetes.”

Sadda continued: “In addition to increased accessibility, a prompt diagnosis made possible with AI means identifying those at risk of blindness and getting them in front of an ophthalmologist for treatment before it is too late.”

Source:

https://aao.org/newsroom/news-releases/detail/using-ai-to-screen-diabetic-eye-disease-feasible

What is diabetic retinopathy?

Retinopathy can progressively develop in diabetic individuals, especially those with poor blood glucose control. Consistently elevated blood glucose can cause damage to the blood vessels at the back of the eye. Small bulges that protrude from damaged blood vessels can leak fluid and blood into the retina, which can impair vision.  

How is retinopathy diagnosed?

Diabetic retinopathy is commonly diagnosed following routine, comprehensive, dilated eye screening.