AI to help close the Indigenous eye care gap
A new artificial intelligence tool developed at CERA aims to identify people at risk of eye disease and provide diagnoses and referral for treatment on the spot.
Indigenous Australians are almost four times more likely to develop diabetes. One in 10 Indigenous Australians will develop vision-threatening diabetic retinopathy.
In 2019, CERA began a 12-month trial with The Fred Hollows Foundation and the Nganampa Health Council in seven remote Indigenous communities in Central Australia’s Anangu Pitjantjatjara Yankunytjatjara (APY) Lands.
As part of the trial a local eye health nurse is using the AI tool to conduct eye tests on 250 people. A sophisticated computer program reviews retinal images taken by a typical fundus camera. It identifies and grades eye diseases within seconds.
Dr Jane Scheetz, who led the study with Professor Mingguang He, says the AI tool’s algorithm was developed and tested over five years using more than 200 000 images of the back of the eye.
These images included common blinding eye diseases from different ethnic populations around the world.
Utilising ‘deep learning’, the computer program can process large amounts of data to recognise problems with the back of the eye and make informed decisions on its own.
Before the AI system was introduced, the usual pathway for investigating potential eye disease included sending images to Adelaide for review by an ophthalmologist.
Because the AI system can assess images instantly patients are provided with results on the spot. By offering instant results and immediate referral for treatment, the AI tool may well prove to be a game changer.
“A major problem with diabetic eye disease is that sometimes it’s not noticed until it’s too late,” says Dr Scheetz.
“The hope is that people will also get used to checking in regularly.
“If you can pick people up earlier and consult with them and show them what’s happening, hopefully we will be able to prevent some eye diseases or progression of disease.”
The trial will conclude with interviews with staff involved with the study to discuss experiences and the pros and cons of the new technology.
The next step will be to compare the AI tool against current telemedicine models and measure accuracy, cost-effectiveness, ease of use and patient and clinician acceptance.