Google DeepMind
Through the medical research partnership, Moorfields will investigate whether DeepMind's AI technology can be used to help spot early signs of eye conditions that human eye care experts might miss.
In order to determine whether DeepMind's AI technology is useful for diagnosing eye conditions, Moorfields is applying the company's algorithms to one million anonymous OCT (Optical Coherence Tomography) scans.
The aim is to determine whether the algorithms can learn to spot early signs of age-related macular degeneration and sight loss that occurs as a result of diabetes.
Mustafa Suleyman, Google DeepMind cofounder and head of DeepMind Health, told Business Insider that he wants DeepMind's AI to understand the structure and nature of eye scans "well enough to be able to try to predict in advance which ones indicate that a patient may be at risk to a particular kind of eye disease."
The algorithms that DeepMind builds are known as machine learning algorithms because they have the ability to learn through training without being explicitly programmed.
"What we want to do is train an algorithm to classify a potential diagnosis in an image in the same way that we've trained algorithms to classify, say, chairs or plants in photos, or that we've trained algorithms to perform well in the Atari [game] simulator," said Suleyman. "Essentially, what it's doing is learning that a particular cluster of pixels are correlated with a particular outcome or action or object that we're looking for."
DeepMind hopes that the research will eventually help eye care professionals to make faster and more accurate diagnoses, leading to better treatment for patients living with eye conditions. The six-year-old research startup claims that the right treatment at the right time can prevent people from going blind, adding that up to 98% of sight loss resulting from diabetes can be prevented by early detection and treatment.
Pearse Keane, a consultant ophthalmologist at Moorfields Eye Hospital, said: "I think machine learning will allow for automated diagnosis of OCT image sets and I think that's really exciting because it will allow us to have earlier detection of these diseases."If we can get earlier detection then we can get much earlier intervention and all the evidence shows that will allow us to save the sight of many people with these conditions."
The two conditions that DeepMind and Moorfields are targeting affect millions of people across the world. Age-related macular degeneration is the most common cause of blindness in the
DeepMind is planning to publish the results, algorithms, and methodologies from the partnership in peer-reviewed journals. Suleyman said: "Other researchers can benefit from the insight and also potentially other commercial partners can see that these sorts of algorithms are really successful."
The reference to other commercial partners is an interesting one as Google has not publicly announced that it is planning to sell DeepMind's technology to other companies.
Previous NHS partnership
This is Google DeepMind's second partnership with an NHS organisation.
Earlier this year, Google DeepMind announced that its new healthcare division is working with Royal Free London NHS Foundation Trust on an app called Streams that can help clinicians to detect acute kidney injury (AKI) in patients. However, the Streams app has faced criticism due to the amount of NHS patient data that it draws on. DeepMind is also trialling a clinical task management smartphone app called Hark with the same NHS trust.
When asked about what data-sharing and regulatory approvals DeepMind has acquired for its partnership with Moorfields, Suleyman said: "All of the appropriate approvals have been sought including going through the local research and development office at Moorfields, as well as approvals from the information governance leads."
Google DeepMind employs roughly 250 people in King's Cross, London, and around 30 of them are now working on DeepMind Health.
Suleyman said that Google DeepMind is likely to announce more NHS partnerships in the future, adding that hundreds of doctors and nurses have contacted the company about potential collaborations.