This site is optimised for modern browsers. For the best experience, please use Google Chrome, Mozilla Firefox, or Microsoft Edge.

Case study: How artificial intelligence is revolutionising ophthalmology

NIHR Clinician Scientist Awards have now been replaced by NIHR Advanced Fellowships. The NIHR Fellowship Programme supports individuals on their trajectory to becoming future leaders in NIHR research.

An idea

In 2015 ophthalmologist Dr Pearse Keane read a magazine interview with a scientist working for DeepMind, a UK based artificial intelligence (AI) company. A LinkedIn message led to a chat over coffee, and a year later a research collaboration between Moorfields Eye Hospital and DeepMind was formed.
But why would an ophthalmologist be interested in working with an AI company?
Pearse was at the start of his NIHR Clinician Scientist Award to assess binocular optical coherence tomography (OCT), having recently completed his NIHR Clinical Lectureship. OCT was a new medical imaging device, similar to an ultrasound, which produces high resolution images of the back of an eye. These scans allow clinicians to diagnose conditions such as wet age-related macular degeneration (AMD), which is the single biggest cause of blindness in the UK. An early diagnosis means treatment can be given that saves sight.
“What drives me is the fact that some people lose sight because they can’t get seen and treated by an eye doctor quickly enough.”
However, thousands of OCT scans are undertaken each week at Moorfields alone, which provides a huge amount of data. These scans need to be analysed quickly, in order to understand which patients need treatment soonest. OCT scans also take place in high street opticians, and so in 2016 Moorfields had 7000 urgent wet AMD referrals, yet only 800 of these patients actually had wet AMD. So could AI detect wet AMD to allow doctors to prioritise the most at risk patients?

Developing an algorithm

This research collaboration involved clinical input from ophthalmologists at Moorfields, scientific input from researchers at UCL, and computer science and engineering expertise from DeepMind. Within two years, the collaborators had developed an algorithm that could identify signs of eye disease using historic, anonymised eye scans. The research, published in Nature Medicine, showed that the AI system could recommend the correct referral decision to the same accuracy as world-leading eye experts for a number of different eye diseases. This was a massive breakthrough.
The next step was to determine if AI could analyse healthy eye scans to predict the development of AMD within 6 months in patients that had already developed it in their other eye. Patients receive treatment for AMD every 4–6 weeks and so OCT scans of both eyes are taken at regular intervals, which provided the researchers with a unique position. The AI system that was developed was found to perform as well as or better than clinicians. This research was published in Nature Medicine in May 2020.
“I’m really excited about the potential of artificial intelligence for healthcare and I believe that ophthalmology can lead the way in this regard.”

Next steps

The team aim to get this system implemented in clinical practice in the coming years. This will require further clinical validation, potentially in prospective clinical studies, as well as receiving regulatory and payor approvals. Pearse and the team at Moorfields are currently working with Google Health, as well as with ophthalmic imaging vendors, to make this a reality.
“In the next 3-5 years we hope to develop the algorithm so that it can be used by healthcare professionals in millions of patients per year, all around the world. We believe that this will facilitate much earlier detection of sight-threatening diseases such as age-related macular degeneration (AMD) and diabetic retinopathy (DR), allowing earlier treatment and potentially saving sight.”
Read more about this career development award on NIHR Funding and Awards.
Read more making a difference stories.
Photo credit: Moorfields Eye Hospital NHS Foundation Trust