Study of use of DeepMind technology at NHS’s flagship eye hospital suggests artificial intelligence could help lead to swifter diagnosis and treatment of eye conditions
For patients with eye diseases, artificial intelligence technology can make the correct care-referral decisions with the same level of success as the very best human doctors, research has found.
The study, which was conducted by Moorfields Eye Hospital NHS Foundation Trust, the University College London Institute of Ophthalmology, and Google-owned AI firm DeepMind Health, used data from about one million “historic de-personalised eye scans”. These were supplied to DeepMind, “along with some related anonymous information about eye condition and disease management provided by Moorfields”.
The images, which covered more than 50 eye conditions, were processed by DeepMind’s AI system, which was tasked with detecting signs of disease and recommending how patients should be referred for further care.
- NHS grants Google access to 1.6 million patient records
- Government names DeepMind chief as top AI adviser
- Google’s DeepMind and NHS restart data-sharing deal with greater transparency
The technology made the appropriate referral decision 94% of the time, which Moorfields claimed is on a par with the success rate of “world-leading eye experts”. The full research paper has been published by medical journal Nature Medicine.
Moorfields hopes that using the technology could ultimately expedite the diagnosis and treatment process – particularly for the most serious eye diseases.
Dr Pearse Keane, consultant ophthalmologist at Moorfields and NIHR clinician scientist at the UCL Institute of Ophthalmology said: “The number of eye scans we’re performing is growing at a pace much faster than human experts are able to interpret them. There is a risk that this may cause delays in the diagnosis and treatment of sight-threatening diseases, which can be devastating for patients.”
He added: “The AI technology we’re developing is designed to prioritise patients who need to be seen and treated urgently by a doctor or eye-care professional. If we can diagnose and treat eye conditions early, it gives us the best chance of saving people’s sight. With further research it could lead to greater consistency and quality of care for patients with eye problems in the future.”
Moorfields claimed the AI system has “been developed with two unique features which maximise its potential use in eye care”.
The first of these is that the technology can provide eye-health professionals with information – including visual representations of eye disease – designed to explain how the system reaches its referral decisions.
The second key feature, Moorfields said, is that DeepMind Health is designed to work with various types of eye scanner – not just the model whose data was used to train the AI system.
“This could significantly increase the number of people who benefit from this technology and future-proof it, so it can still be used even as… scanners are upgraded or replaced over time,” the hospital said.
Clinical trials of the use of AI in eye-care patient referrals will now be conducted. If these trials are successful, the technology can seek regulatory approval for use in hospitals.
If and when such approval is achieved, Moorfields has the right to use the technology for free across its 30-strong UK network of hospitals and clinics, for a period of up to five years.
Moorfields first entered into a partnership with DeepMind two years ago. As pair of the engagement, the Google-owned firm has “invested significant resources to clean, curate and label Moorfields’ de-identified research dataset to create one of the most advanced eye research databases in the world”, the hospital said.
This database – which is still owned by Moorfields as a non-commercial public asset – is providing information for nine further research studies.
A year ago, the Information Commissioner’s Office ruled that the Royal Free NHS Trust in north London had breached the Data Protection Act during an engagement with DeepMind in which details of 1.6 million patients were shared with the AI firm. The ICO found that there were “several shortcomings in how the data was handled”, including a lack of adequate information and opt-out opportunities for patients whose data was involved in the trial.