Google's AI product detects retinal diseases with unprecedented accuracy
American Academy of Ophthalmology News Aug 23, 2018
DeepMind plans to unleash an artificial intelligence (AI) product capable of diagnosing 50 ophthalmic conditions from a simple OCT scan, the company reports. The technology detected signs of retinal disease more accurately than human experts in a study published August 13 in Nature Medicine.
“These are early results, but they show that our system could handle the wide variety of patients found in routine clinical practice,” the authors posted on their website. “In the long-term, we hope this will help doctors quickly prioritize patients who need urgent treatment—which could ultimately save sight.”
The product stems from a collaboration between Moorfields Eye Hospital, University College London Institute of Ophthalmology, and London-based DeepMind. Under a 5-year agreement that began in 2016, Moorfields pledged to share 1 million anonymous fundus images and OCT scans with DeepMind, along with associated details about the eye condition and disease management.
In their recent study, the team created the AI architecture and trained it on 14,884 retinal OCT scans, including those from patients with diabetic retinopathy, AMD, and 48 other common conditions. Upon testing with an additional 997 images, their findings show that the technology can not only identify eye disease with a 94% accuracy, it can offer referral suggestions on par with those of retinal experts and explain how it arrived at that particular decision. According to the team, these features will drastically cut down the typically lengthy delays between scan and treatment.
While the AI was trained on Topcon OCT scans, researchers say their algorithm works on a broad range of OCT devices. This will allow the technology to be applied in clinics across the globe without hardware restrictions.
The company told Bloomberg they plan to test the technology in humans next year. If the product passes its road test, DeepMind plans to seek regulatory approval and provide the technology to UK practitioners at no charge for the first 5 years.
“Detecting eye diseases as early as possible gives patients the best possible chance of getting the right treatments. I really believe that one day this work will be a great benefit to patients across the National Health Service,” says Mustafa Suleyman, co-founder of DeepMind.
-
Exclusive Write-ups & Webinars by KOLs
-
Daily Quiz by specialty
-
Paid Market Research Surveys
-
Case discussions, News & Journals' summaries