Article: Researchers eye health benefit from AI | 21 June 2021
Elsewhere in the area of AI diagnostics, Three Springs Technology has been working closely with WA-based health care company Resonance Health on tools associated with liver disease: FerriSmart and HepaFat-AI.
Three Springs works with businesses such as Resonance Health to develop technology and bring it to market. In the case of FerriScan, data gathered is applied to train AI for the quantification of liver iron concentration.
As a result, it is claimed to be the world’s first and only regulatory cleared AI tool for LIC, having secured approval by the Food and Drug Administration in the US.
“These solutions are being used by clinicians all over the world,” Three Springs director James Telders said. “In addition, we market Aquila 3.0, a next-generation, secure and cloud-based radiology workflow management software with access to integrated AI tools.”
Half of Three Springs’ staff are devoted to its own R&D pipeline of AI solutions in the medical space, closely collaborating with research institutions in Australia and abroad.
Founded in 2016 by Sherief Khorshid, Three Springs began as an AI consulting business working on projects in the mining, oil and gas and medical sectors.
In 2019, it merged with ThinkCDS, a competing WA business run by data scientist Mike Clark.
“I joined the combined business in that same year,” Mr Telders told Business News.
“Since 2020, the three partners have made the strategic decision to focus solely on medical innovations, and we see this field as an important opportunity where we will continue to build on the successes the company has achieved.”
Self-supervised learning is the future of medical AI, according to Mr Clark. “Right now, AI is limited by a need for large amounts of human examples,” Mr Clark said.
“We want it to save clinicians’ time, but to do that it requires they do a lot of work creating examples. This is a big limiting factor, especially for busy doctors.
“The solution is self-supervised learning, in particular contrastive learning, which only requires a hundredth as many examples.”
He said every AI project had a lifecycle that must be followed: from acquiring, analysing and cleaning data, to developing and finessing the model. Regulatory clearance followed, with most projects taking at least a year from start to proof of concept.
A new product must have a clinical need, be technologically feasible, and fit the regulatory pathway. Only when it ticked these boxes did a company start thinking about commercial viability.
Three Springs is one of only a handful of WA companies to build commercially successful, FDA cleared medical software using AI.
The following is an extract, subscribers can view the full Business News article here.