13 March 2024
The consultant ophthalmologist and professor of artificial medical intelligence shared that ophthalmology is the busiest specialty within the NHS – with around 10 million appointments each year.
“We are drowning in the number of people who we need to see within the hospital eye service,” Keane shared.
“There are some people who go blind because of delays in being seen and treated,” he emphasised.
He shared his belief that AI has the potential to transform healthcare – while ophthalmology could be an exemplar of this shift.
Keane noted that the first AI system to receive Food and Drug Administration (FDA) clearance in 2018 was IDX – a diagnostic system for the detection of diabetic retinopathy.
He shared how Google has developed an AI system, called ARDA (Automated Retinal Disease Assessment), that has been used to screen more than 200,000 patients for diabetic retinopathy in India and Thailand.
Keane outlined how there would be challenges in implementing a similar system within the UK context.
“We already have advanced human-based screening. There are quite a few processes that need to be satisfied to ensure that if we introduce a system it is better than what is already in place,” he said.
Keane noted that one of the most prestigious international medical journals, the New England Journal of Medicine, created an offshoot dedicated specifically to AI research in November 2023.
He added that to date more than 500 medical AI systems have received FDA clearance.
Keane discussed a collaboration between Moorfields Eye Hospital and DeepMind that developed an algorithm capable of identifying retinal disease with the same level of accuracy as experienced ophthalmologists.
“The idea is to bring the expertise of a retinal specialist into the community,” he shared.
After a ground-breaking study was published in Nature Medicine in 2018, Keane acknowledged hurdles in translating this research into a service that patients receive.
“We got a lot of hype around that, but here we are – six years later – and this AI system is not in real-world use,” Keane shared.
“That speaks to some of the challenges around implementing these systems,” he added.
There are some people who go blind because of delays in being seen and treated
Diverse applications
Keane highlighted the broad range of roles that AI can play within healthcare.
“I think that the most important application is not going to be in direct patient care but in clinical trial planning and recruitment,” he said.
He noted that one of the common reasons that clinical trials fail is because of flaws in the recruitment process.
Keane shared that an AI system could be used to identify patients within a clinic who might be eligible for clinical trial recruitment.
Following the approval of different treatments, AI could also be used to identify which patients would be suitable for treatment.
“These are all applications of AI that don’t necessarily require regulatory approval but could still bring immense value to patients,” he said.
Keane outlined a project where the ophthalmic images of Moorfields Eye Hospital patients were linked to the Hospital Episode Statistics database.
The images were from patients over the age of 40 and covered a 10-year period.
A series of papers have been published as a result of this research, including a paper in March 2023 in JAMA Psychiatry that explored the relationship between retinal features and schizophrenia.
In August 2023, a paper was published in Neurology that examined retinal changes in Parkinson’s disease.
“There were subtle changes in the retina on average seven years before patients were diagnosed with Parkinson’s disease,” Keane highlighted.
Keane emphasised the potential for ophthalmic imaging to provide insight on systemic health conditions by reaching a broad group of people in the community.
He shared that only around 9% of people take up the health check that is offered once they turn 40 – compared to 80% of people who present for a sight test at least once every four years.
Comments (0)
You must be logged in to join the discussion. Log in