Google DeepMind and Moorfields Eye Hospital have joined forces to apply artificial intelligence to analyse more than a million eye scans in an attempt to detect early signs of sight loss.
The British artificial intelligence wing of Google will attempt to use machine learning to spot early signs of common eye conditions, such as macular degeneration and diabetic retinopathy, which doctors might miss.
The number of people suffering from sight loss in the UK is set to double from 2 to nearly 4 million by 2050, but many cases are preventable. Up to 98% of sight loss resulting from diabetes can be prevented by early detection and treatment.
“I had the brainwave that deep learning could be really good at looking at the images of the eye.”
The collaboration marks the first time DeepMind has applied machine learning to a healthcare project. The idea came to Pearse Keane, a consultant ophthalmologist at Moorfields, after he read a story about a DeepMind machine being trained to play Atari games.
“I’d been reading about deep learning and the success that technology had had in image recognition,” Keane told the Guardian. “I had the brainwave that deep learning could be really good at looking at the images of the eye.”
If successful, the technology will be capable of detecting eye conditions with a single digital scan.
Mustafa Suleyman, a DeepMind co-founder, said: “We set up DeepMind because we wanted to use AI to help solve some of society's biggest challenges, and diabetic retinopathy is the fastest growing cause of blindness worldwide. There are more than 350m sufferers across the planet. I really believe that one day this work will be a great benefit to patients across the NHS.”
The Moorfields collaboration is the second partnership between DeepMind and the NHS. In April, the health service courted controversy when it was revealed that the Royal Free NHS Trust had shared 1.6 million patients’ potentially identifiable medical records with DeepMind.
While the NHS regularly shares its data with third parties, Morrfields' announcement is almost certainly aimed to avoid any further criticism over data-sharing. In addition the Royal Free patient records were potentially identifiable, Moorfields’ are not, meaning the barriers for securing privacy are considerably lower. The ICO does not require consent for non-identifiable data sharing. It should also be noted that current Moorfields patients’ data will not be used in the partnership.
The latest collaboration is the first in which DeepMind is using artificial intelligence purely for medical research. The earlier partnership with the Royal Free NHS Trust enabled the monitoring of kidney functions using a smartphone app.
Story by Oscar Williams