The rate of diagnosis of diabetic retinopathy is very important. Delays, in this case, are fraught with complete blindness. An algorithm, designed in one of the Indian clinics by Google and Verily experts, is supposed to facilitate the work.
Diabetic retinopathy is a serious complication of diabetes mellitus that affects the eyeball vessels. It manifests itself in every ninth patient and can really cause blindness.
The staff of the Aravind Ophthalmological Clinic, located in the city of Madurai, has been working on the technology of automatic detection of this pathology for 16 years. But only two years ago, they managed to achieve acceptable speed and accuracy in collaboration with Google engineers.
The algorithm was taught to diagnose diabetic retinopathy on 128 thousand ophthalmologic images of eyeballs. The effectiveness of the method was tested on 2 sets of 12 thousand images each. The accuracy of the diagnosis reached 97.5 percent. It is faster and more accurate than doctors.
A similar study was conducted by Google and the Rajavithi Hospital in Thailand last year. The results of that algorithm were similar: the accuracy reached 95 percent, whereas human ophthalmologists were wrong in every fourth case.
The Google algorithm has recently received a sign of compliance with the standards of the European Union CE. Thanks to this algorithm, the doctors of the clinic managed to diagnose more patients and prescribe the necessary treatment.
