Computers have been taught how to detect eye disease in a “jaw-dropping” breakthrough that could save many people from going blind.
A pioneering trial at Moorfields Eye Hospital found an artificial intelligence system was as effective as world-leading experts in spotting disease.
Technology created by Google’s AI company DeepMind diagnosed problems with 94 per cent accuracy within seconds in more than 50 types of eye disease.
Crucially, in about 1,000 tests, not a single case of disease requiring urgent referral was missed. Doctors hope to begin clinical trials next year and believe the AI system could become standard hospital practice within three years.
This could shorten waiting times for assessment by a consultant from weeks to days, prioritising those most at risk before irreversible damage occurs.
It would also offer speedy reassurance to the thousands of patients who fear for their sight after a “false positive” scan at a high street optician.
DeepMind believes the same technology could be used to analyse X-ray, CT and MRI scans, widening its potential use to cancer and a host of other diseases.
Pearse Keane, a consultant ophthalmologist at Moorfields and scientist at the UCL Institute of Ophthalmology, said the results, published today in the journal Nature Medicine, were “jaw-dropping”. He told the Evening Standard: “They are pretty stunning results and could, I think, transform the whole speciality of ophthalmology in the next few years.”
The project used 14,884 historic and anonymised 3D, high-definition OCT (optical coherence tomography) retinal scans to “teach” the computer how to spot 10 features of eye disease.
A further 1,000 scans from Moorfields patients were then analysed by the computer and by eight of the hospital’s consultant ophthalmologists and specialist optometrists, and the results compared.
Both were asked to decide whether to recommend the patient for urgent referral, semi-urgent referral, routine referral or for observation. The computer even indicated what had prompted it to make a certain diagnosis.
Dr Keane said: “There was no case where somebody had something serious and the algorithm recommended they just be observed. That was the most important thing.
“There were only two cases out of 1,000 where the algorithm had a ‘false positive’ — where the patient had nothing wrong with them but the algorithm recommended urgent or semi-urgent.”
More than two million Britons have some form of sight loss. Eye disease is one of the biggest causes, but in about 80 per cent of cases sight loss can be prevented by early detection and treatment. Clinicians are struggling to check an increasing number of scans, creating the risk of delayed diagnosis. NHS rules require high-risk patients to be seen and treated within two weeks.
The AI system’s backers believe it can overcome “bottlenecks” in analysing scans and will ensure patients at greatest risk are seen first. Dr Keane said: “This is not going to replace consultants. No one is going to be having an injection in their eye or surgery purely on the basis of the machine. It’s just about getting people in front of the appropriate specialist at the earliest possible point.”
The Moorfields study began in 2016 to investigate whether AI could improve the care of patients with sight-threatening diseases such as age-related macular degeneration and diabetic eye disease. Macular disease is the biggest cause of sight loss in the UK, affecting 600,000 people.
The AI system looks for symptoms such as haemorrhages, lesions and fluid build-up. It has the ability to adapt to different types of scanner, enabling widespread use and future-proofing it against advances in technology.
Mustafa Suleyman, the British co-founder of DeepMind Health, said: “These incredibly exciting results … could, in time, transform the diagnosis, treatment and management of patients with sight-threatening eye conditions around the world.”
Matt Hancock, the Health Secretary, said: “This is hugely exciting and exactly the type of technology which will benefit the NHS in the long term.”