Artificial intelligence

A major artificial intelligence (AI) project is working on creating a general algorithm for the diagnosis of retinal disease and could eventually reinvent the eye exam, reported Pearse Keane MD, Consultant Ophthalmologist at Moorfields Eye Hospital, UK, at the Irish College of Ophthalmologists 2017 Annual Conference. Dr Keane gave a fascinating presentation on reinventing the eye exam in the era of big data and AI, through his work on Google’s DeepMind AI project. Mr Keane is working with DeepMind on a wide range of macular diseases, including age-related macular degeneration (AMD), diabetic macular oedema, retinal vein occlusion among many others. The project aims to investigate how deep learning could help analyse optical coherence tomography (OCT) scans and make the diagnosis of these conditions more efficient and effective, leading to earlier detection and intervention for patients. As part of the research project between Moorfields Eye Hospital and DeepMind, deep learning is being applied to one million anonymous historic eye scans, to look for early signs of eye conditions that humans might miss. While current OCT imaging is very effective, these images contain a huge amount of complex data, which takes considerable time, training and experience to analyse correctly, with very large volumes of such scans now taken daily in busy ophthalmology clinics, Mr Keane noted. The idea of using AI is that it is a technique that the algorithm learns from experience so it can look at a process of thousands, or millions, of scans and become as good as a retina specialist at diagnosing these conditions. “It is important to emphasise that this is a research collaboration and we are hoping to publish a research publication sometime before the end of 2017 showing a ‘proof of concept’ of the algorithm, though it will be a little bit further in the future before it can be used in practice,” Mr Keane told EuroTimes. The project has the potential to revolutionise the way eye exams are carried out, which could lead to earlier detection and treatment of common eye diseases, and Mr Keane’s presentation attracted considerable interest at the conference. Meanwhile, Mr Keane also cautioned against the expansion of OCT, traditionally a hospital- and specialist clinic-based technology, into high-street optometrist chains unless appropriate training is in place for optometrists and/or artificial intelligence systems have been developed, saying that such widespread and casual availability could end up overwhelming ophthalmology clinics with needless patient referrals. Pearse Keane: pearse.keane@moorfields.nhs.uk

Tags: retina