Google AI being trained to spot diabetic blindness


Artificial Intelligence has seen exponential growth recently, thanks to the major tech companies spending heavily in AI research. Google is one of the foremost innovators in this space. One of their latest use of AI is to catch diabetic blindness. Google’s deep learning AI is being trained to catch the spread of diabetic retinopathy. There are 415 million diabetics across the world at risk. Many of them are in the developing world and lack sufficient health care access to treat it.


The disease is diagnosed by examining a picture of the back of the eye. The Google team tackling this have developed a dataset of 128,000 images. These images have been examined by 3-7 ophthalmologists from a panel of 54. The ophthalmologists then marked damaged areas of the eye and fed it back into Google’s machine learning system. Thanks to this, Google was able to build a highly reliable diagnostic tool. They tested 12,000 images and the diagnostic made by the AI was on-par with the ophthalmologists.

The future of this system is envisioned to be able to diagnose the disease from complex 3D images as opposed to the 2D photographs used now. The team is currently looking into automating this diagnostic process in order to better serve patients in remote locations. Google is still away from widespread use of their technology. Studies using larger clinical groups and FDA approvals are yet to come.