DeepMind, a UK-based subsidiary of Google, today announced a new research project with National Health Service specialist Moorfields Hospital to use artificial intelligence to detect and treat blindness-causing eye diseases.
The project is DeepMind’s second collaboration with NHS, although earlier projects did not use AI. Using a million anonymous eye scans, the research partnership will investigate how machine learning could help analyze eye scans known as Optical Coherence Tomography (OCT) images, creating algorithms that can detect early warning signs of two particular eye diseases: wet age-related macular degeneration and diabetic retinopathy. The goal of the project is that eventually the AI will be able to recognize such conditions with just a digital scan.
The collaboration came about when Moorfields opthalmalogist Dr. Pearse Keane contacted DeepMind, looking to find ways to work together to tackle the two specific eye diseases, which affect over 100 million people worldwide, with diabetic retinopathy as the fastest-growing cause of blindness.
The information gleaned from the research will not directly impact current patient care – Moorfields is sharing with DeepMind the scans it has collected over the years through routine patient care – but will help researchers develop faster ways to detect and intervene with sight deterioration. Currently, eye care professionals use digital scans of the back of the eyes along with OCTs, which are complicated and time-consuming to analyze, and traditional computer analysis tools have been insufficient in reading them.
Google acquired DeepMind in January 2014. In February of this year, Google DeepMind announced partnerships with the UK’s National Health Services to develop medical apps. One app was Streams, which gave doctors information about their acute kidney failure patients, and the other was Hark, which helps doctors and nurses organize information that is currently managed with hand-written notes.
The company ran into controversy with the earlier NHS app-development programs, in which DeepMind had access to the full care history of some 1.6 million patients. This project shouldn't raise any patient security concerns, though: the set of anonymous eyes scans, and some related information about eye condition and disease management, has been collected by Moorfields over time through routine care.
“This means it’s not possible to identify any individual patients from the scans,” Google DeepMind said in a statement. “And they’re also historic scans, meaning that while the results of our research may be used to improve future care, they won’t affect the care any patient receives today.”