Machine learning and artificial intelligence hold promise for medical researchers from around the globe hoping to identify and cure diseases, and one of the first targets for the nascent technology is diabetic retinopathy, which can cause blindness in people with diabetes.
Google's DeepMind project is training AI algorithms to identify patterns in eye scans that may indicate the disease's presence, something that takes human specialists a long time to do.
"It takes me my whole life experience to follow one patient's history," opthamologist Peng Tee Khaw told The Guardian. "And yet patients rely on my experience to predict their future. If we could use machine-assisted deep learning, we could be so much better at doing this, because then I could have the experience of 10,000 lifetimes."
Khaw is the director of ophthalmology research at Moorfields Eye Hospital in east London, which recently announced a partnership with DeepMind to use historical scans of patients' eyes to train computers how to detect diabetic retinopathy.
The goal is to use AI to reduce the time it takes to diagnose the disease: as many as 50 percent of people who do not receive timely treatment will be registered as blind within five years, according to DeepMind. Machine-learning algorithms would automatically detect and segment eye scans into patients who are at risk and those who are not.
The DeepMind research mirrors a recent effort in California, where Stanford researchers trained algorithms to detect eye diseases using images from a VA hospital in Palo Alto. That project used a hybrid approach to reduce demands on computer-processing power, according to Apaar Sadhwani, a Stanford Ph.D. candidate who presented the team's research at Nvidia's GPU Technology Conference in April. They performed large-scale searches with millions of parameters on low-resolution images, and used a neural network to process a few high-definition ones.
The stakes are high. Even in developed countries like the US or the UK, there are not enough specialists to expeditiously detect the disease. More than 3,000 eye scans are performed every week at Moorfields Eye Hospital alone. And it's not just diabetic retinopathy.
"Our research with DeepMind has the potential to revolutionise the way professionals carry out eye tests and could lead to earlier detection and treatment of common eye diseases such as age-related macular degeneration," Khaw said in a statement.