The Australian company that wants to help one million people a day

The chief executive of healthcare technology firm Harrison.ai has moved to dismiss privacy concerns about his start-up, describing them as a misunderstanding as it plots an expansion into the lucrative US market.

Harrison.ai gives radiologists and pathologists access to AI technology so they can more efficiently and accurately scan X-rays for cancers and illnesses. The Sydney start-up has already raised more than $150 million to pursue its stated goal of helping one million people a day by 2025.

Dimitry (left) and Aengus Tran. Credit: Louie Douvis

An investigation by online publication Crikey has alleged that Harrison.ai trained its flagship product, Annalise.ai, using scans of potentially hundreds of thousands of Australians obtained by radiology provider I-MED seemingly without express consent from patients.

Speaking in an interview with this masthead, Aengus Tran, who founded the company in 2018 with his brother Dimitry, said those concerns were a “misunderstanding” and that his start-up anonymises patient data to the extent it cannot be re-identified.

“I think, from the outside, maybe people made an assumption that we are processing personal information,” Aengus Tran said.

“A picture of your face is very different to a chest X-ray and a diagnostic report that has been anonymised and de-identified. We have a really robust anonymisation and data-protection pipeline, where data is completely stripped of personal information, and [we have] gone through quite great lengths to ensure that it cannot be re-identified.

“That helps us meet the requirement of the Privacy Act and therefore enables our use of data to be fully permissible within the bounds of that act. The Privacy Act actually makes it very clear that if you thoroughly anonymise and de-identify the patient, data is no longer considered personal information.”

I-MED has also published a statement calling the reports “inaccurate”. “I-MED de-identified data using best practice frameworks developed by the CSIRO and the Office of the Australian Information Commissioner,” it said.

Anxiety and discussion around how artificial intelligence technologies treat customer privacy has spiked since Australia’s privacy commissioner, Carly Kind, found that retailer Bunnings had breached privacy laws with facial recognition systems that it used to combat crime.

Privacy Commissioner Carly Kind has launched a preliminary inquiry into I-MED.Credit: Dion Georgopoulos

Kind has launched preliminary inquiries to determine if I-MED complied with the Privacy Act through its partnership with Harrison.ai.

A patient’s medical scans are among the most sensitive data types. At the core of the issue is whether scans from I-MED were truly “de-identified” when they were used for training Harrison.ai’s models. Aengus Tran says they were.

A spokeswoman for the privacy commissioner’s office said its preliminary inquiries were ongoing. “I-MED has been co-operative, and we are reviewing information received from the company,” she said.

“One of the key issues we are considering is whether or not personal information was de-identified,” the spokeswoman said. “Generally speaking, entities should be aware that de-identification is context-dependent and may be difficult to achieve. In addition, entities seeking to use de-identified information to train generative AI models should be aware that de-identifying personal information is a use of the personal information for a secondary purpose.”

Aengus Tran said that Harrison.ai’s technology is in 131 British hospitals and now used by one in two radiologists in Australia. The start-up has grown by 300 per cent year-on-year, and Tran said the time is right to enter the US market.

Harrison.ai co-founders Dimitry and Aengus Tran have done deals with Virtus Health and I-MED.Credit: George Fetting

“We’ve been waiting for the right opportunity to do that,” he said. “And recently that opportunity arrived.”

In October, Harrison.ai was awarded Medicare reimbursement by CMS, the US federal agency that administers the Medicare program. It means US hospitals will be paid up to $US240 a scan processed by Harrison.ai’s technology.

“Alongside our 12 FDA clearances that we’ve had in the last 18 months, we feel this is the right time for an Australian company punching well above its weight to crack the American market,” Tran said.

“We very much believe in keeping the centre of gravity, especially on the development side, in Australia, but we know we need to grow internationally to make the global impact we need to make.

“The United States is about 50 per cent of the world’s radiology market, by volume, so we’ve established ourselves as a clear winner in the rest of the world, and we now see a doubling of our total opportunities.”

Aengus’ brother Dimitry, who is Harrison.ai’s other co-founder and its deputy chief executive, has moved to Boston to establish the start-up’s US base. The pair originally migrated to Australia from Vietnam for high school.

“The US is half of the global radiology market,” Dimitry Tran said. “Building on our customer momentum in the UK, Asia and Australia, we’re in a great position to bring our technology to the US and win the market.

“Relocating to the US signifies how important the market is to us ... The timing is perfect, as we make the most of our recently awarded Medicare reimbursement – our first for the US market.”

Aengus said that 2025 was expected to include a significant capital raising, likely north of $100 million. The company previously raised $29 million in late 2019, and another $129 million two years later. The start-up’s backers include Blackbird Ventures, Hong Kong’s Horizons Ventures, Scott Farquhar and Kim Jackson’s Skip Capital and ASX-listed hospitals giant Ramsay Health Care.

Pathology giant Sonic Healthcare and radiology provider I-MED are also investors, and Tesla chair Robyn Denholm sits on its board.

“Up to this point, we’ve been building our AI technology in the radiology domain, and to add to that, next year we’ll be launching our AI technology in pathology,” Tran said.

“This has been the result of our collaboration with Sonic Healthcare, and next year we’ll start launching products and commercialising that for the first time.

“It’s very exciting. Radiology is very large, and there’s a road map of products that we are looking to develop across CT [scans of] chest, abdomen, pelvis and X-rays of the musculoskeletal system. So, essentially, trying to complete our AI portfolio across radiology.

“The nice thing about this is that we have a proven formula. You could say that we’ve built the espresso-making machines, and we make a nice cup of coffee, and now the plan is to put more beans in and finish the job.”

The Business Briefing newsletter delivers major stories, exclusive coverage and expert opinion. Sign up to get it every weekday morning.