Coronavirus: Australia leads the world in AI-assisted coronavirus diagnosis tool

Artificial intelligence-driven technology designed to train clinicians to diagnose breast cancer is being adapted for COVID-19 treatment.

Health Minister Greg Hunt says the use of CT scans will quickly and more accurately diagnose the severity of coronavirus in patients who are having difficulty breathing. Picture: Gary Ramage

Artificial intelligence-driven technology designed to help train clinicians to diagnose breast cancer is being adapted for COVID-19 treatment under a new initiative supporting frontline health workers.

Federal Health Minister Greg Hunt will announce a $3m package on Monday to support new web-based technology to help ­clinicians “rapidly upgrade their skills in diagnosing and managing patients with severe coronavirus through CT scans”.

Patrick Brennan, who specialises in diagnostic imaging at the University of Sydney and is chief executive of the DetectED-X start-up, has been developing the technology for the past eight years.

Professor Brennan began to adapt the technology for the coronavirus about five weeks ago, leading to the creation of CovED, an online training module collating positive cases of coronavirus diagnosed through CT scans.

Mr Hunt allocated $1m to the CovED initiative, from the Medical Research Future Fund, and $2m for the Australian Partnership for Preparedness Research on Infectious Disease Emergencies to “understand the prevalence, improve diagnostic tests and develop innovative tools that limit transmission of coronavirus”.

“The CovED initiative involves using artificial intelligence to support frontline health workers, using CT scans to quickly and more accurately diagnose the severity of coronavirus in patients who are having difficulty breathing,” Mr Hunt said. “This project will be delivered through a consortium led by DetectED-X and the coronavirus Image Biobank, with support from iCoreLab and a large consortium of University of Sydney and clinical experts.”

Professor Brennan said the training module asked clinicians to diagnose cases from CT scans that had already been assessed by experts, with the clinicians alerted to what mistakes they made. The AI technology learns the mistakes the user makes and generates further learning modules specifically to address the individual clinician’s previous errors.

“Clinicians, wherever they are in the world, look at real imaging cases that we know the truth of, which have got disease and which hasn’t. We get them to diagnose, and the algorithm assesses how good they are, which cases were right, assesses their performance and what their weakness was. It identifies each individual error,” he said.

“The algorithm then compares what the clinicians thought was abnormal or normal with what the experts thought was abnormal or normal.”

In the 10 days since launching CovED for free online it has been accessed by frontline clinicians across Australia, as well as in 77 countries around the world.

“We wanted to get it out there and wanted people using it as quickly as possible,” Professor Brennan said. “We’re pretty convinced this will transform people’s abilities to recognise the severity and progression of COVID-19.

“If you’re a clinician in Cyprus, you will just log onto our website and register for free, and the first thing you see is these COVID-19 cases our experts have judged.

“And then we have our experts through the platform and algorithm that can let that Cypriot clinician know how good their diagnosis is, and then create a ­lesson specifically to address the skills they need to perfect.”

He said the funding boost would assist DetectED-X in developing new AI technology to help clinicians around the world when they encounter “a difficult case”.