Google’s Artificial Intelligences can do a lot of things. From identifying faces, to holding conversations, to beating humans in Go and Chess — Google has an AI for every situation. In fact, I am just waiting for the day when the company creates an AI that starts preaching a new religion. However, physical wellbeing comes before mental and Google’s new Artificial Intelligence may just be the first concrete step along the path to AI driven health-care.
Working in collaboration with leading doctors and medical practitioners, Google has created an AI that is able to automatically identify diabetic retinopathy. Diabetic retinopathy, in case you are unaware of it, is a complication brought about by diabetes and affects eyes. While in its initial stages diabetic retinopathy may cause no symptoms or only mild vision problems, it can worsen and indeed, it is one of the chief causes of blindness in adults.
By deploying the same deep learning technology that enables it to recognize the content of a picture and if it contains faces, animals, and objects — Google’s new AI can detect diabetic retinopathy by examining retinal photos. Awesome right? Well, it gets even better. According to a paper published today in the Journal of the American Medical Association, the AI was touted to have succeeded at almost the same rate as human opthamologists.
Speaking on the topic, Lily Peng, Physician and biomedical engineer at Google said,
We were able to take something core to Google—classifying cats and dogs and faces—and apply it to another sort of problem.
To be clear, Google is not attempting to replace Opthamologists here. By automating the process of screening, more and more humans will be screened and as such, the rate of retinal blindness arising out of diabetes can be brought down. Indeed, this project can trace its roots to India, where a native Google researcher realized that this technology could be used to combat the issues arising out of limited healthcare availability.
Speaking on the topic, David McColloch, clinical professor of medicine at the University of Washington said,
In many places, doctors are already using photos to diagnose the condition without seeing patients in person. “This is a well validated technology that can bring screening services to remote locations where diabetic retinal eye screening is less available.
And Google’s AI is capable of doing just that — except of course, it can process the images faster and can do it with greater precision by utilizing machine learning and data techniques.
As with every other project that uses AI, the team fed the Opthamologist AI — for want of a better word — was fed almost 128,000 retinal images into its neural network. Each of these images was first reviewed by at least three doctors and the inputs also provided to the AI. All this effort finally culminated into a system that could recognize and identify the images with a surprising degree of accuracy.
As per Google, the system was able to avoid false negatives and false positives with 90 percent of the time. This is above and over the 80 percent accuracy prescribed by the National Institutes of Health for diabetic retinopathy screens.
Well, this literally is awesome. Our own creations are now joining hands to take care of their masters. Compared to the potential that all these new technologies actually have — Google’s retina disease diagnosing AI is but a small step. However, it does offer us a very deep look at just what is possible if we decide to harness the power of the technologies we have developed, for the greater good.