“Now we’re just testing it in the real world,” says Prof Turner. “Here’s a patient in front of us, here are three different AIs to road-test. How does it go, practically speaking, with this new camera, the new on-the-spot diagnosis, and what difference does it make for that patient on the day?
“It was a research trial. The patient signed a consent form to say they were happy with all this fancy AI stuff. And then we asked them: Do you like this on-the-spot facility? Does it help you? And they said they absolutely loved it. So, the patients were happy with the experience, and the AI was extremely accurate, as good as or better than the doctors.”
In the end, Google Health’s advanced diagnostic tool called ARDA (Automated Retinal Disease Assessment) was the clear winner of the three AI models. This technology has now been handed over to Ninox Vision to provide the service in Australia, so it has no further connection to Google.
The camera takes a photo of the patient’s retina to diagnose diabetic retinopathy at the back of the eye, and then the digital image is immediately transferred to the cloud – where it’s analysed by the fully integrated AI software that provides an instant diagnosis, rather than the patient having to wait perhaps weeks for the results.
“AI is the only reason we can now do that on-the-spot diagnosis,” Prof Turner explains. “The machine securely transfers the photo to the cloud, where AI does its magic to work out whether diabetes is affecting the eye or not. And that’s been validated as being very accurate for Aboriginal people.”
The technology is also easy to master after just a 10-minute training session.
“So, my goal,” Prof Turner says, “is to say to GP practices: if you have a health assistant or a receptionist, they might be able to take the photo on the way in before the patient has even seen anyone because they don’t need any eye drops. And by the time the patient is seeing the doctor, the report is already in the medical record.”
