The company reveals that the disease is curable, but only if diagnosed on time.
Diabetic retinopathy can seriously harm eyesight and lead to blindness.
Google has trained its artificial intelligence to recognize the symptoms of this disease in its earliest stages without the need for medical examinations and an ophthalmologist’s opinion.
The diagnostic method includes screening of the eyeball's bottom. The deep learning algorithm used a database of 128 000 individual images, each analyzed by 3 to 7 ophthalmologists.
The doctors identified the damaged areas and added them to the deep learning system. Based on this information, Google has built an algorithm that is extremely accurate in diagnosing the disease.
12 000 tests were carried out, and artificial intelligence accuracy results were no worse, even slightly better, that ophthalmologists’ results.
The team working on this project hopes to train its AI to detect the disease using 3D images instead of the 2D ones used so far.
Google aims to make the diagnostic process automatic so that it can be used by inexperienced medical staff in remote areas.