Associate Professor Angus Turner screening an Indigenous patient. Photo courtesy of Lions Eye Institute.
Lions Outback Vision founder and director, Associate Professor Angus Turner, has won a prestigious award for partnering with Google to use artificial intelligence to detect diabetic retinopathy (DR) in Indigenous Australians.
His team found the AI system performance was as good as, if not better, than retinal specialists at accurately detecting and grading DR in Aboriginal patients in Western Australia.
Research Australia awarded Assoc Prof Turner its inaugural Digital Health Technology Award in its 2023 national Health and Medical Research Awards on 2 November. The new category recognises innovators who are at the forefront of digital health technologies.
AI has not previously been used for routine DR screening in Australia. Assoc Prof Turner validated Google’s AI-based application Automated Retinal Disease Assessment (ARDA) software for DR detection in Indigenous Australians. The system integrates into retinal cameras and works in real time with an operator taking photos of the eyes, uploading them and a machine learning algorithm interpreting scans, giving immediate results.
AI has not previously been used for routine DR screening in Australia
ARDA was created after Google trained the AI model to read eye images by asking ophthalmologists to manually review over 100,000 de-identified retinal scans, grading them from one to five, healthy to diseased. These scans trained the algorithm in image recognition so the AI could predict signs of disease.
Vision Loss 14 Times More Common
Diabetic-induced vision loss is 14 times more common in Indigenous than non-Indigenous Australians. Assoc Prof Turner said Lions Outback Vision at the Lions Eye Institute had been looking for ways to bridge this gap and through the collaboration, began researching AI to make screening accessible, efficient and enable earlier detection.
As part of the partnership, the team conducted a study with Perth’s Derbarl Yerrigan Health Service to validate the model.
“Our study shows that a deep learning system can detect DR in an Indigenous Australian cohort with improved sensitivity and similar specificity compared with a retina specialist,” Assoc Prof Turner and colleagues said in the British Journal of Ophthalmology (in February.)
He told Google Australia: “The paper found that the machine learning model performs on-par with a retinal specialist. Validating the model across diverse patient populations is critical to determine the role and potential of this technology in a clinical setting and can help to reduce racial bias in AI models.”
Lions Outback Vision is now delivering the screening in Pilbara communities in a customised van, improving screening access and preventing blindness.
Research Australia CEO Nadia Levin commended Assoc Prof Turner for bringing the technology to remote communities, saying it would save sight and lives.
the machine learning model performs on-par with a retinal specialist
Assoc Prof Turner said the work had implications for detecting other issues and helping stop patients falling through the cracks.
“The eye can also tell us about the kidneys and the heart, because we can see the veins and arteries in the eye,” he said. “Our solution means we will be in a unique position to trial other deep learning systems in a real-world environment, using retinal photos to determine risk of other health issues such as anaemia, cardiovascular disease and chronic kidney disease.
“AI has immense promise but we need to measure its impact and make sure we’re deploying something that actually works.”