Diabetic retinopathy detection now possible with Artificial Intelligence

Diabetic retinopathy is the fastest growing cause of blindness, with nearly 415 million diabetic patients at risk worldwide. All individuals with diabetes whether it is Type 1 or Type 2, need to undergo regular and repetitive annual retinal screening for early detection and timely treatment of diabetic retinopathy. If caught early, the disease can be treated otherwise, it can lead to irreversible blindness.

The best way to detect diabetic eye disease is to have a specialist examine images of the back of the eye and then grade them for disease presence and severity. Severity is determined by the type of lesions present, which are indicative of bleeding and fluid leakage in the eye. Interpreting these photographs requires specialized training,  unfortunately in many parts of the world, where diabetes is prevalent, medical specialists capable of detecting the disease are not available. This leads to many diabetic retinopathy cases going unnoticed, finally leading to loss of vision.

image

AI comes in action

Google decided to leverage Machine learning  to compact this problem. They made this possible by employing Deep learning technique which directly graded the disease stage by scanning the macula-centered retinal fundus images.  This new method of screening increased efficiency, reproducibility, and coverage of screening programs and made available access to doctors to only those patients who were really needing medical attention and thus reduced complications from Diabetic retinopathy. The tool was developed using a reference set of  128 175 images, of  which 33 246 had referable diabetic retinopathy. The images were obtained from EyePACS in the United States and 3 eye hospitals in India (Aravind Eye Hospital, Sankara Nethralaya, and Narayana Nethralaya). 

FDA recently approved first artificial intelligence (AI) device capable of screening for diabetic retinopathy (DR) developed by Iowa-based biotech firm IDx, LLC. This software is designed for use by primary care physicians and other health providers who treat patients with diabetes. This software screens adult patients without requiring an eye specialist to scan images for signs of DR. Images of the patient’s retina are captured with the help of a Topcon NW400 camera and then uploaded to a cloud server. After checking the images the software gives out one of two possible results: More than mild diabetic retinopathy detected hence refer to an eye care professional or negative for more than mild diabetic retinopathy, so rescreen in next 12 months.

Diabetes mellitus and its complications including diabetic retinopathy has become one of the 21st century’s major health challenges, but WHO believes that 80% of all visual impairment caused by diabetic retinopathy can be cured with regular screening and early detection using AI driven tools.