Artificial intelligence (AI) has made remarkable strides in the field of healthcare, and its latest foray into ophthalmology is poised to revolutionize the way we detect and treat eye diseases. Researchers have unveiled an AI system, known as RETFound, which not only identifies sight-threatening eye conditions but also holds the potential to predict the onset of various other health conditions. Developed using millions of eye scans from the UK’s National Health Service (NHS), this innovative AI model has the power to transform healthcare on a global scale. In this article, we explore the groundbreaking potential of RETFound and its implications for the future of medical diagnosis and treatment.
A pioneering AI foundation model
At the heart of this groundbreaking development is RETFound, the first AI foundation model specifically tailored for ophthalmology. Foundation models like RETFound are vast, complex AI systems trained on extensive datasets without labels. They can subsequently be fine-tuned for a wide range of specific tasks. RETFound stands out by consistently outperforming existing AI systems across complex clinical tasks. Moreover, it addresses a significant limitation of many current AI models by demonstrating proficiency in diverse populations and even among patients with rare diseases.
The global impact of RETFound
The significance of RETFound goes beyond its technical capabilities. The research team behind this remarkable AI system has decided to make it freely available to institutions worldwide. This generous move aims to propel global efforts in the detection and treatment of blindness using AI. By sharing their advancements with the international community, the researchers hope to accelerate progress in eye care and broaden the scope of AI-driven healthcare.
Expanding diagnostic horizons
RETFound has the potential to enhance the diagnosis of some of the most debilitating eye diseases, including diabetic retinopathy and glaucoma. Beyond eye-related conditions, it holds the promise of predicting systemic diseases such as Parkinson’s, stroke, and heart failure. The concept of identifying general health issues through the eyes is part of an emerging field known as ‘oculomics,’ a term coined in 2020 by co-author Professor Alastair Denniston from the Institute of Inflammation and Ageing. This intersection of ophthalmology and general health prediction could revolutionize preventive medicine.
Efficiency through self-supervised learning
One of the key challenges in developing AI models is the acquisition of expert human labels, a process that is often expensive and time-consuming. RETFound addresses this challenge by achieving comparable performance to other AI systems while using as little as 10 percent of human labels in its dataset. This efficiency boost is made possible through an innovative self-supervising approach. In this method, RETFound masks parts of an image and then learns to predict the missing portions autonomously. This breakthrough in label efficiency streamlines the development process and reduces the resource-intensive aspects of AI model training.
The data backbone
To train RETFound effectively, the research team curated a dataset comprising a staggering 1.6 million images from Moorfields Eye Hospital. Leveraging the cutting-edge AI tools and infrastructure provided by INSIGHT, the NHS-led health data research hub for eye health, the team harnessed the power of AI in healthcare. INSIGHT, led by Moorfields Eye Hospital NHS Foundation Trust in partnership with University Hospitals Birmingham NHS Foundation Trust, evolved from a pioneering 2016 collaboration with DeepMind, now known as Google DeepMind. This collaboration laid the foundation for the development of powerful computing and AI capabilities in healthcare.
A blueprint for AI-enabled healthcare
Professor Pearse Keane, the senior author of the study, envisions a bright future for AI-enabled healthcare in the UK and beyond. He believes that by combining high-quality clinical data from the NHS with top-tier computer science expertise from universities, the UK can position itself as a world leader in AI-enabled healthcare. The work on RETFound serves as a template for achieving this ambitious goal, demonstrating how collaboration between healthcare institutions and technology experts can drive innovation and improve patient outcomes.
RETFound’s emergence as a pioneering AI foundation model in ophthalmology represents a significant leap forward in healthcare. Its potential to revolutionize eye disease diagnosis and predict broader health conditions underscores the transformative power of AI in medicine. With the global community benefiting from this innovation, the future of healthcare holds the promise of earlier, more accurate diagnoses and improved patient care, all thanks to the visionary integration of artificial intelligence.
