Intelligent screening for diabetic retinopathy developed
Local scientists and medical experts have jointly developed an AI system for the quick and accurate identification of diabetic retinopathy, a major eye complication of diabetes.
An article titled “A Deep Learning System for Detecting Diabetic Retinopathy Across the Disease Spectrum” has been published by world-leading journal Nature Communications.
The research is based on the world’s largest database for eye-ground images and established the intelligent diabetic retinopathy assisted diagnosis system. This can conduct automatic diagnosis of different stages of the disease and do real-time judgement on the quality of eye-ground images, according to Dr Jia Weiping from Shanghai 6th People’s Hospital and one of the leading expert of the research.
Diabetic retinopathy has no symptoms in the early stages. When patients have vision problems, the disease is already very serious and can develop into lifelong vision damage and even blindness.
“All diabetic patients should receive eye-ground image screening regularly, however there are 130 million diabetics in China with only 44,800 eye doctors,” Jia said. “Doctors must work on the problem for a long time and need to be experienced.”
To solve the problem and meet the demand for diabetic retinopathy screening, experts developed the intelligent system DeepDR, which is included in Shanghai’s three-year public health project.
The system was established after studying nearly 700,000 eye-ground pictures. It can identify patients automatically and offer suggestion to doctors, greatly relieving workloads and diagnosis difficulties for grassroots doctors. It has been used by many domestic medical facilities as an effective screening tool for diabetic retinopathy.
In addition to China, some 40 countries and regions have also used the system, which offers Chinese-featured intelligent management plan for world diabetes prevention and control.