Google is poised to begin a grand experiment in using machine learning to widen access to healthcare. If it is successful, it could see the company help protect millions of people with diabetes from an eye disease that leads to blindness.

Last year researchers at the search and ads company announced that they had trained image recognition algorithms to detect signs of diabetes-related eye disease roughly as well as human experts. The software examines photos of a patient’s retina to spot tiny aneurisms indicating the early stages of a condition called diabetic retinopathy, which causes blindness if untreated.

At the 2017 WIRED Business Conference in New York City today, a leader of Google’s project said that work has begun on integrating the technology into a chain of eye hospitals in India. The country is one of the many places around the world where a lack of ophthalmologists means many diabetics don’t get the recommended annual screening for diabetic retinopathy, said Lily Peng, a product manager with the Google Brain AI research group.

“This kind of blindness is completely preventable, but because people can’t get screened, half suffer vision loss before they’re detected,” she said, describing the current situation in India. “One of the promises of this technology is being able to make healthcare more accessible.” There are more than 400 million people worldwide with diabetes, including 70 million in India.

Peng, who is an MD, was featured on WIRED’s Next List of 20 tech visionaries creating the future earlier this year.

In India, Google is working with the Aravind Eye Care System, a network of eye hospitals established in the late 1970s and credited with helping reduce the incidence of blindness caused by cataracts in the country. Aravind helped Google develop its retinal screening system by contributing some of the images needed to train its image parsing algorithms. The system uses the same deep learning technique that allows Google’s image search and photo storage service do things like differentiate between dogs, cats, and people.

Google’s paper last year just described the accuracy of that technology when applied to retinal images, not its use in the clinic. Peng said today that Google has just finished a clinical study in India—meaning the technology was used in real patient care—with Aravind. Work is now under way on getting the technology into routine use with patients, she said.

Peng dismissed suggestions that while this technology might be good for patients, it could mean fewer jobs for doctors. She said Google’s algorithms would instead do screening work not being done today due to skills shortages while freeing physicians for more important tasks. “There’s not enough expertise to go, we need to have our specialists working on treating people who are sick,” said Peng.



Wired