The new system could be used not only to identify eye conditions but also to give insight into risk of heart attacks, stroke and Parkinson’s disease.
RETFound, the first AI foundation model in ophthalmology, was developed using millions of eye scans from the NHS. The research team are making the system freely available to use by any institution worldwide, to forward global efforts to detect and treat blindness using AI. This work has been published in Nature.
A foundation model describes a very large, complex AI system, trained on huge amounts of unlabelled data, which can be fine-tuned for a diverse range of subsequent tasks. RETFound consistently outperforms existing state-of-the-art AI systems across a range of complex clinical tasks, and even more importantly, it addresses a significant shortcoming of many current AI systems by working well in diverse populations, and in patients with rare disease.
Professor Pearse Keane, senior author, from University College London said: “This is another big step towards using AI to reinvent the eye examination for the 21st century, both in the UK and globally. We show several exemplar conditions where RETFound can be used, but it has the potential to be developed further for hundreds of other sight-threatening eye diseases that we haven’t yet explored.”
RETFound could help improve diagnosis of some of the most debilitating eye diseases, including diabetic retinopathy and glaucoma, and predict systemic diseases such as Parkinson’s, stroke and heart failure. Identifying general health issues through the eyes is an emerging science called ‘oculomics’ – a term coined in 2020 by co-author Professor Alastair Denniston, from the Institute of Inflammation and Ageing.
The eye offers a ‘window’ to examine our overall health. Non-invasive measurements we obtain from the eye can be used to give health practitioners an insight into non-eye related complex diseases and other problems associated with ageing.
Professor Alastair Denniston.
One of the key challenges when developing AI models is the need for expert human labels, which are often expensive and time-consuming to acquire. As demonstrated in the paper, RETFound is able to match the performance of other AI systems whilst using as little as 10 percent of human labels in its dataset. This improvement in label efficiency is achieved by using an innovative self-supervising approach in which RETFound masks parts of an image, and then learns to predict the missing portions by itself.
RETFound was trained with a curated dataset of 1.6 million images from Moorfields Eye Hospital. This used AI tools and infrastructure provided by INSIGHT, the NHS-led health data research hub for eye health. INSIGHT is led by Moorfields Eye Hospital NHS Foundation Trust working in partnership with University Hospitals Birmingham NHS Foundation Trust. The hub’s powerful computing and AI capabilities evolved from a 2016 research collaboration with DeepMind, now Google DeepMind.
Professor Keane notes, “If the UK can combine high quality clinical data from the NHS, with top computer science expertise from its universities, it has the true potential to be a world leader in AI-enabled healthcare. We believe that our work provides a template for how this can be done.”
