Artificial intelligence will determine the eye disease photo

Ddiabeticheskaya retinopathy is one of the most insidious complications of diabetes. The disease affects the blood vessels of the retina and increases the risk of blindness 25 times. For the diagnosis of retinopathy patients need every year to pass inspection: the measurement of visual acuity and ophthalmoscopy. The procedure is not complicated, but modern technology can make it even more simple and convenient. Specialists of the research laboratory of Google Brain Team trained artificial intelligence (AI) to identify diabetic retinopathy for a photo with an accuracy of 90%.

According to who, diabetes today affects more than 422 million people. Each of them is in high risk of developing diabetic retinopathy. The disease can be successfully treated at an early stage, but if you delay the diagnosis, you can completely lose vision.

Associated diseases:

DepressionCancer Alzheimer’s DiseaseDementiaBlindness Diabetic retinopathy Diabetes

Diabetic retinopathy: traditional diagnosis

The traditional method of diagnosis of diabetic retinopathy ophthalmoscopy. Using special instruments, the doctor examines the eye and estimates the condition of the blood vessels. About the disease is evidenced by the expansion and deformation of veins, dilatation and increased permeability of capillaries, formation of microaneurysms, nerve fiber necrosis and other injuries.

A qualified doctor can detect diabetic retinopathy on fundus photographs, but there are experts in all areas. Develop Google Brain Team will facilitate diagnosis for patients with limited access to health services.

A database of 128 thousand of 2D photographs, which were taught AI, collected Indian ophthalmologists. To minimize the possibility of error, each photo is checked at least 3 specialists. Then the images were narguile in the neural network, and deep machine learning has helped the AI to interpret the ocular fundus images to diagnose the disease with 90 percent accuracy.

In the future, scientists want to teach the AI to recognize the pathology of various retinal layers on 3D-imaging for the diagnosis of other eye diseases.

Recall that artificial intelligence is already used for early diagnosis of dementia, Alzheimer’s, cancer, depression and some other diseases. The software is able to make it indispensable for physicians of different profiles.