Five ‘rights’ for AI and the eye

Artificial intelligence already proving useful in retina diagnostics

Default banner image for Five ‘rights’ for AI and the eye
Photo of Leigh Spielberg

Daniel Ting,MD
"Artificial intelligence represents the fourth industrial revolution, and its application in ophthalmology has clearly begun,” said Daniel Ting, MD, Singapore National Eye Centre, Singapore at EURETINA 2020 Virtual. Dr Ting brought viewers along for a ride through a world of technological developments that most ophthalmologists barely know even exist. Dr Ting and his team have already published two major review papers on the topic, on both artificial intelligence (AI) and deep learning.“In these papers, we outline what AI can already achieve with deep learning based on fundus imaging for conditions like diabetic retinopathy, glaucoma suspects, AMD and retinopathy of prematurity,” he said. “It is also becoming increasingly evident that AI can see things that we as ophthalmologists cannot see,” said Dr Ting. Based on deep learning, AI software can determine both the age and the gender of a patient based on a simple fundus image of the posterior pole. More usefully, papers have been published regarding the prediction of cardiovascular risk factors from retinal fundus photographs via deep learning. But what about for patients with eye disease? “Even more relevant for practising retinal specialists, AI has been shown to be able to predict conversion from dry to wet AMD, better than retinal specialists.” Okay, AI machine learning sounds great. But how does one get started? Dr Ting said: “You have to start with the ‘Five Rights’. You have to ask the right questions, identify the right clinical data, identify the right technical partner, understand the right concepts and identify the right enabler who can implement and commercialise the final product.” Most important for ophthalmologists and researchers working in the field, asking the right question involves identifying an unmet clinical need, such as screening large populations for diabetic retinopathy. So how will AI and deep learning get a chance at acceptance by physicians? “It seems that the current pandemic crisis has stimulated more widespread acceptance of AI,” he said, referring to an article describing how physicians, faced with staff shortages and overwhelming patient loads, are starting to use artificial intelligence to triage COVID-19 patients. “Once they’ve proven their worth, these tools may be here to stay,” concluded Dr Ting. “We know that the patients, physicians, regulators, providers, funders are becoming more willing to fund some of these technologies in the healthcare setting,” he said, before quoting Sir Winston Churchill: “Never let a good crisis go to waste.” In the recent Expertscape ranking (2010 – 2021), Dr Ting is also ranked 1st on deep learning (among 35K candidates) across all medical and technical domains.  Daniel Ting: daniel.ting.s.w@singhealth.com.sg