Australia carries a grim distinction that shapes almost everything about how it thinks about skin: it has one of the highest rates of skin cancer on the planet, and roughly two in three Australians will be diagnosed with some form of it before they turn 70. Against that backdrop, a group of Adelaide researchers has now put forward an artificial intelligence system designed to improve how skin diseases are diagnosed, a development that speaks directly to a public health problem the country has never quite managed to get on top of.
The work, reported by Glam Adelaide, centres on a machine-learning tool trained to recognise patterns across a wide range of skin conditions, from common rashes and lesions through to the early signs of melanoma and non-melanoma cancers. The pitch is familiar in medical AI circles but no less significant for it: an algorithm that can look at an image of a patient’s skin and flag, quickly and consistently, whether something warrants closer attention from a clinician.
Why Adelaide, and why now
South Australia has quietly become one of the country’s more active hubs for health and medical AI, with university research groups, hospital networks and a growing cluster of medtech startups all working on tools that sit alongside clinicians rather than replacing them. The state government has leaned into the theme as well, having established a royal commission examining artificial intelligence, a move that put Adelaide at the centre of a national conversation about how far and how fast the technology should be allowed into public life.
Skin disease is, in many ways, an obvious place to start. Diagnosis relies heavily on visual assessment, which makes it well suited to image-based machine learning. The volume of cases is enormous, dermatologists are in short supply, and the gap between a benign mole and a dangerous melanoma can come down to subtle features that are easy to miss and slow to confirm. A system that helps triage the flood of cases, sorting the urgent from the routine, could ease pressure on a stretched specialist workforce and, in the best case, catch cancers earlier when they are far more treatable.
The promise, and the caveats
Supporters of this kind of technology argue that the real prize is not replacing the dermatologist but extending their reach. A GP in a regional town, a pharmacist, or even a nurse in a remote clinic could use an AI tool to decide who needs to be referred urgently and who can safely wait, which matters enormously in a country where the nearest specialist can be several hundred kilometres and many weeks away. In that framing, the Adelaide system is less a diagnostic oracle than a triage assistant, one that could shorten the path from a worrying spot on the skin to a proper clinical assessment.
The cautions are just as well rehearsed, and they are worth taking seriously. Medical AI systems are only as good as the data they are trained on, and dermatology algorithms have a documented history of performing less well on darker skin tones, largely because the images used to build them have skewed heavily towards lighter skin. In a multicultural nation, and one with a First Nations population whose health outcomes already lag, a tool that works beautifully on some patients and poorly on others is not a neutral piece of software. Independent validation, transparency about where the training images came from, and clear limits on how the tool is used will all matter before anything reaches a clinic.
There is also the question of regulation. Any product that claims to diagnose disease is likely to be treated as a medical device by the Therapeutic Goods Administration, which means a research prototype, however promising, still has a long road ahead before it can be deployed at scale. Clinicians tend to stress that these systems should support a human decision rather than make it, and that liability, consent and the handling of sensitive patient images need to be nailed down early rather than patched on later.
What it means for Australia
For a country that spends a fortune on skin cancer, the potential upside is hard to overstate. Skin cancers cost the health system well over a billion dollars a year to treat, and the burden falls unevenly on outdoor workers, older Australians and those living outside the major cities. Anything that pushes detection earlier, reduces unnecessary specialist referrals, or brings a credible first assessment to communities that currently go without could pay for itself many times over, both in dollars and in lives.
It also fits a broader pattern taking shape across Australian healthcare, where AI is arriving fastest in the image-heavy specialties. Radiology and cardiology firms listed on the ASX have already built businesses around algorithms that read scans, and clinical documentation tools that transcribe consultations have produced some of the country’s most valuable health startups. Dermatology is a logical next frontier, and an Adelaide-grown system would add to a growing body of Australian intellectual property in a field that has often been dominated by overseas players.
The catch, as ever, is trust. Australians have shown they will embrace health technology when it is convenient and clearly beneficial, but confidence in AI more broadly remains fragile, and a single high-profile failure in something as consequential as cancer diagnosis could set the whole field back. The researchers behind this work will know that the science is only half the battle; the other half is proving, to regulators, clinicians and patients alike, that the tool is safe, fair and genuinely useful in the messy conditions of a real clinic rather than the controlled environment of a lab.
What is next
The immediate path runs through the usual milestones of medical research: larger and more diverse datasets, peer-reviewed publication, and clinical trials that test the system against the judgement of experienced dermatologists on real patients. If those steps hold up, the more interesting questions become practical ones. Where does the tool sit in the patient journey, who is allowed to use it, and how is its output checked? Adelaide has positioned itself as a place willing to grapple with exactly those questions, and a home-grown diagnostic tool gives the state a concrete test case for the kind of careful, clinically grounded AI it says it wants to build.
For now, the work sits where most promising medical AI does, at the hopeful end of the research pipeline rather than the bedside. But in a country that leads the world in skin cancer, even an incremental improvement in how quickly and accurately those cancers are spotted is the sort of advance worth watching closely.
Sources: Glam Adelaide.
