ASX health imaging stocks view AI as a growth tool, not a disruptor. Pic: Getty Images
ASX health imaging firms see AI as an opportunity, not a threat, with growth potential intact
Pro Medicus caught up in tech sell-off as investors question impact of AI
Morgans says AI is a powerful tool for workflow efficiency but won’t replace radiology or enterprise-grade imaging
Artificial intelligence (AI) has become a major talking point for investors and analysts in recent months, contributing to volatility across global software stocks as markets weigh the technology’s potential to disrupt the sector.
The rise of powerful AI coding tools from companies such as Anthropic has fuelled speculation that software development could become faster, cheaper and easier to replicate, raising questions about the durability of traditional SaaS business models.
Pro Medicus (ASX:PME) has long been one of the standout growth stories on both the S&P/ASX 200 Index and the S&P/ASX 200 Health Care Index.
With a market cap of ~$14.5 billion, the company is the largest of ASX-listed medical imaging software names and has delivered strong earnings growth for many years, supported by global demand for its Visage imaging platform.
Like many tech companies, however, the share price of Pro Medicus has been volatile recently as investors question valuations and the potential impact of AI. Despite reporting a strong first-half result, the stock has fallen roughly 40% year-to-date.
Pro Medicus software can’t be readily replicated
Both CEO and co-founder Sam Hupert and Morgans healthcare analyst Iain Wilkie, who covers Pro Medicus, argue the concerns may be overstated.
In commentary released to the market, Hupert said Pro Medicus had been caught up in broader tech sector sentiment driven by massive capital spending on AI infrastructure.
“There are concerns at the huge level of capital expenditure committed to develop AI and build data centres to run it,” Hupert said.
“Ours is a capital-light, software-only model. If anything, we will be the beneficiaries of the infrastructure funded by others.”
Hupert also pushed back on the idea that new AI tools could easily replicate complex industry-grade software.
“This, in our view, is an overly simplistic generalisation, one that certainly doesn’t apply to us,” he said.
Hupert said Visage 7 was built from the ground up using proprietary technology.
“It is a very specialised, highly technical, patented suite of software that incorporates more than 30 years of domain knowledge,” he said.
“It is not a product that can be readily replicated with or without AI. We have not left a roadmap for others to follow.”
Thirdly, Hupert said their offering was more than software but was all the systems and methods built around it enabling Pro Medicus to deliver the “highly sophisticated, deeply integrated solution used by many of the world’s leading medical institutions”.
Still a need for radiologists
Hupert said the view that AI could disrupt the diagnostic imaging market to the point where there won’t be a need for radiologists was an “oversimplification and overestimation of the current capability” in the sector.
“Based on recent manpower surveys it is estimated that the efficiencies that AI currently delivers, and is predicted to deliver soon, will be counterbalanced by the increasing workload generated by new forms of imaging and larger data sets,” he said.
“In other words, it will be used to ‘play catchup’ rather than replace radiologists who are currently in huge demand.”
Wilkie said radiology was structurally constrained and had been for a decade with CT and MRI volumes growing faster than radiologist supply, which means bottlenecks, reporting queues, and burnout.
“AI is absolutely a powerful tool for workflow efficiency, but it doesn’t replace radiology, and crucially, it doesn’t replace the enterprise-grade imaging backbone these tools depend on,” he said.
“AI still requires robust routing, data management, visualisation, audit trails, and compliance layers – which is exactly where PME plays. They’re the rails the system runs on.”
Robust regulatory environment
Wilkie said radiology information systems (RIS) and enterprise imaging workflows were among the most heavily regulated categories in healthcare IT.
“In the US, HIPAA mandates strict privacy, encryption, audit trails and role-based access,” he said.
“In Europe, GDPR governs every piece of personal health data, and the MDR, along with the incoming EU AI Act, classifies AI enabled radiology tools as high risk, requiring human oversight, conformity assessments and post-market surveillance.
The FDA treats RIS and associated workflow software as Class II medical devices subject to full quality system regulation.
“In Australia, the TGA regulates anything used for serious disease diagnosis.”
He said globally, hospitals can’t even accredit without meeting standards set by medical bodies like the American College of Radiology.
“All of this means clinical infrastructure can’t be swapped out quickly or cheaply.
“Procurement cycles stretch for years – we see it with Pro Medicus and Mach7, where contracts can take years before implementation.
“Healthcare institutions are deeply risk averse. Any AI first challenger must navigate security, explainability, liability and reimbursement hurdles that are orders of magnitude higher than anything consumer AI faces.”
Improving AI may strengthen Singular Health’s position
Singular Health Group (ASX:SHG) CEO and director Denning Chong said AI already played a role in the company’s 3DICOM technology and was likely to become more important over time.
However, he said Singular Health was building infrastructure around imaging workflows rather than a single AI algorithm.
“3DICOM is really a workflow and interoperability platform across patient access, education and clinical use – not just a single algorithm,” Chong said.
He said improving AI tools could strengthen the company’s position.
“As AI improves and more FDA-regulated models become available, it becomes more valuable to have a compliant platform where those tools can be delivered into clinical workflows,” he said.
“Our marketplace is designed to allow multiple imaging AI applications to operate through a single secure infrastructure.”
Solving barriers in healthcare imaging systems
Singular Health is developing technology to address a long-standing problem in how medical images are accessed, visualised and shared across healthcare systems.
Medical images are typically stored in a format known as DICOM and managed through Picture Archiving and Communication Systems (PACS), which often operate independently across hospitals and healthcare providers.
The fragmentation can make it difficult to share imaging records and frequently leads to duplicate scans, a problem estimated to add tens of billions of dollars to annual healthcare costs in the US.
3DICOM is designed to act as a bridge between these systems. It combines three core technologies including a Medical File Transfer Protocol, a Volumetric Rendering Platform and AI in the Cloud – enabling clinicians to securely transfer, view and analyse imaging data.
Chong said the bigger challenge in healthcare was not developing algorithms, but securely moving and managing large volumes of imaging data between systems.
“The harder problem is secure interoperability and regulated deployment – integrating with PACS systems, transferring large imaging datasets and meeting privacy and regulatory requirements,” he said.
“Singular sits on the secure rails that imaging data moves across. As AI adoption increases, the value of that infrastructure layer should increase as well.”
AI aiding development for EMVision
Developer of a new category of portable brain scanners EMvision Medical Devices (ASX:EMV) CEO and managing director Scott Kirkland told Stockhead AI was helping accelerate development rather than posing a threat to its technology.
EMVision is undertaking a pivotal validation trial for its first commercial device, the emu, across several leading academic medical centres. The portable brain scanner is designed to rapidly detect and classify strokes at the patient’s bedside.
The company’s second-generation device, First Responder, is a lighter, backpack-sized version of the emu designed for use in ambulances or at the scene of a neurological emergency.
First Responder is designed to diagnose stroke and stroke type at the scene and, in the future, traumatic brain injury (TBI) with pre-hospital road and air studies also underway for this device.
Generating entirely new datasets
Kirkland said most common AI solutions in the healthcare sector were trained on large datasets of existing medical images such as CT, X-ray, ultrasound or MRI scans.
“Because those datasets are widely available, there are relatively few barriers for well-resourced AI companies to train and validate algorithms on them,” he said.
“Which means it can be a quicker path to market for these solutions, but it can also make it very difficult for a pure-play AI solution to build a competitive moat that can be defended over the longer term.”
Kirkland said EMVision was generating entirely new device-specific clinical dataset, that it trains and validates its own AI models on, collected via its brain scanners, which use electromagnetic signals rather than conventional CT or MRI techniques.
“We’re gathering unique electromagnetic scattering measurements of the brain, and those datasets simply don’t exist in the public domain,” he said.
Generating entirely new datasets creates several layers of defensibility for EMVision.
“What we have is a physical moat in the form of proprietary hardware, a data moat through exclusive datasets generated by that hardware, and an algorithm moat because we train our own AI models on that data to produce diagnostic outputs. All of which is protected by a robust IP portfolio,” Kirkland said.
“Our edge is really owning the hardware, the data and the AI models.”
Kirkland said rather than posing a competitive threat if anything AI significantly sped up development.
“It certainly accelerates development cycles on many fronts, from advanced simulations to testing and debugging code, which historically required many engineers to execute,” he said.
“In addition, the signals being sent and received by our hardware are incredibly complex, and AI does an exceptional job of interpreting those signals to deliver a highly reliable diagnostic result at the point-of-care.”
The views, information, or opinions expressed in the interview in this article are solely those of the interviewee and do not represent the views of Stockhead.
Stockhead has not provided, endorsed or otherwise assumed responsibility for any financial product advice contained in this article.
At Stockhead, we tell it like it is. While Singular Health and EMVision Medical Systems are Stockhead advertisers, the companies did not sponsor this article.