For years, researchers at Moorfields Eye Hospital and the UCL Institute of Ophthalmology have been working to use their world-leading expertise in artificial intelligence (AI) to improve eye care.
But earlier this year, led by Dr Pearse Keane and working with international collaborators, they set out on a different tack: to see whether they could diagnose common eye diseases (like age-related macular degeneration (AMD) and glaucoma) using automated machine learning - effectively getting AI to build new AI for them. Today, they have published their findings in The Lancet Digital Health – a journal that promotes the use of digital technologies in healthcare to improve outcomes for patients across the world.
"AI that builds AI"
Moorfields Eye Hospital is the largest eye hospital in Europe, and so generates a huge amount of data on patients’ eyes. Recent advances in AI research - both at Moorfields and the UCL Institute of Ophthalmology and elsewhere – have made it possible to analyse and interpret that data in new ways which can greatly improve the quality of patient care.
However, developing new AI systems that can make sense of all that data takes time, specialised computers and advanced technical skills. Pearse realised that if he could make it easier for those without access to AI expertise and advanced equipment to build their own programs, it would open up a whole new avenue of AI research in healthcare - led by healthcare professionals working in the clinic.
If this technology can be used more widely – in particular by healthcare professionals without computer programming experience – it will really speed up the development of these systems with the potential for significant patient benefits.
Together with Dr Anthony Khawaja, Dr Konstantinos Balaskas and Dr Edward Korot, and supported by a Moorfields Eye Charity springboard award, Pearse has now successfully used the automated machine learning platform Google AutoML to develop a new algorithm for analysing medical images to diagnose eye diseases. This ‘AI built by AI’ can classify eye images just as well as established deep-learning algorithms – at least for simple tasks.
This new approach makes creating and using AI more accessible, and so opens up the possibility for more clinicians to make use of it to provide patients with a faster diagnosis. However, there’s more work to be done before this process can be applied in the clinic, and making sure that AI is used correctly in practice will require AI experts and clinicians to work closely together to develop good regulation and guidelines around it.
These results are the first to come out from Moorfields Eye Charity’s new springboard award programme, designed to support researchers looking to test a new idea which they hope will open up new avenues of research. Ailish Murray, Director of Grants and Research at Moorfields Eye Charity, said:
It’s still early days but these results are really encouraging when we consider their potential to bring clinical benefits to a wider range of patients. We’re excited to see what our continued funding of ground-breaking AI research at Moorfields will lead to next.
Read the full open-access paper on The Lancet Digital Health's website. To find out more about the impact of what we fund, read our recent Impact report.