Transforming Healthcare using AI - Lessons from Ophthalmology


In our Zoom for Thought on July 28th, 2020, UCD Discovery Director Prof Patricia Maguire spoke to Dr Pearse Keane (pictured), Consultant Ophthalmologist at Moorfields Eye Hospital, London and Associate Professor at UCL Institute of Ophthalmology, about, “Transforming Healthcare using Artificial Intelligence - Lessons from Ophthalmology”. In case you missed it, here are our Top Takeaway Thoughts


Groundbreaking Research

Dr Keane’s work has gained prominence in recent years. In 2016 he initiated a collaboration between Moorfields and DeepMind - the research arm of Google - to develop AI algorithms for the early detection and treatment of retinal disease. Using over one million anonymised NHS eye scans they will build a machine learning system to recognise sight-threatening conditions from just a digital scan of the eye. Keane feels passionate about the work, not only because “it’s a chance to publish some groundbreaking research in a journal like Nature Medicine and the like, but also, perhaps even more importantly, to try and see your work translated to be used by millions of people”.


Eye-opening Numbers

In 2017 ophthalmology overtook orthopedics as the number one disease specialty in all of the NHS in terms of clinic appointments. “Nearly 10% of clinic appointments in the NHS are for eyes. That’s nearly ten million appointments per year.” Keane and his colleagues are “drowning in the number of patients we need to see and I’m sure it’s the same in Ireland and in other places around the world”. Unfortunately people are losing sight, sometimes irreversibly, because they cannot be treated quickly enough. “And I think new technologies and innovation and AI will play a role, at least in part, in addressing this problem.” 


Increasing AMD 

A particular problem in Ireland and the UK is age-related macular degeneration (AMD), which can cause blindness. “With an ageing population we’re seeing more of it,” says Keane, quoting research estimating the prevalence of early or intermediate AMD in Europe at 25% of those aged 60 and older. “That’s mind-blowing numbers as the population gets older. Of course, that’s to say nothing of diabetes, all these other eye diseases.” 



Ophthalmology is one of the most technology-driven medical specialities. “In the 1950s they began to use lasers for diabetic retinopathy and lasers to remove your glasses, surgical techniques to remove your cataracts, stem cell gene therapy, et cetera.” Because it is so tech-driven and imaging-driven, ophthalmology is “the perfect substrate for artificial intelligence”. But Keane does not believe AI will replace doctors anytime soon. In 2030 he predicts still being “super-busy” and with the new problem of “information overload” from “ten different types of imaging”, including genomic, proteomic, metabolomic, sensor-data and app-data. “We’ll need to use AI to integrate all of these information streams and make decisions for our patients, and combine that with empathy and wisdom.” 


Data Protection

Though anonymised and retrospective, the sensitivities around sharing NHS data with “a multinational tech company like Google” was “the single most important aspect of the work”. Along with bringing on board all the major eye disease charities and the Royal College of Ophthalmologists, Keane himself has given over 200 public talks about the collaboration in an effort to “do our best in terms of transparency, data protection and just showing people that we’re trying to do the right thing for the right reasons”. 


Academia + Industry = Innovation  

The importance of interdisciplinary research “is the thing that I probably feel most passionately about”. Keane is grateful for a seven-year Future Leaders Fellowship grant from UK Research and Innovation. “It’s a new funding scheme that is specifically aimed at projects that are ambitious and adventurous. It’s flexible funding and it is specifically at the boundary between disciplines, the boundary between industry and academia, and I think that’s where we can really get things done.”



Virtual Progress

Since COVID-19, Moorfields has “moved much more towards virtual and telemedicine consultations” and “clinically, that’s important”. Keane's AI research group has become more practised in collaborating remotely using “data sets in the cloud”, which is “the future in terms of clinical AI research”. He sees this time as “the chance to try to use new technologies and innovation to bring world leading expertise out of the hospital, into the community and into the home. I think that should be the direction of travel for healthcare”.


Calling all Med Grads!

Keane would “love to hear from” Irish graduates and “particularly anyone in UCD”. Aside from their medical qualifications and training, “fundamental skills in statistics” are a bonus for anyone interested in joining his research group, being “the most transferable skills you can possibly have”. He predicts that one of the next “big areas” will be automated deep learning, and most tech companies have such platforms. “I would say just go and explore some of these systems because I think we’re at a stage with AI that’s a bit like the 1970s in terms of computing where we’re going to move towards the democratisation of this technology.” When there are “biomedical scientists, doctors, other healthcare professionals really engaging - that’s when we’ll find thousands of new use cases for this technology”.


This article was brought to you by UCD Institute for Discovery, fuelling interdisciplinary research collaborations in UCD.