Introduction by Croakey: Just as Croakey has been asking how AI can be safely, effectively and equitably deployed in healthcare, more than 250 local and international professionals have converged on Sydney to explore the answers for the future of radiology, cancer care and the broader health system.
Experts in medical AI, diagnostic imaging, radiation oncology and data science gathered at #Intelligence26: AI and the Future of Practice, hosted by the Royal Australian and New Zealand College of Radiologists (RANZCR) from 24–25 July.
Our journalist Marie McInerney attended the first day of the conference for the Croakey Conference News Service, talking with key speakers and convenors about the changes in healthcare’s leading AI field.
See her interviews below for conference highlights, many of which are also relevant for the wider healthcare system: from excitement about the promise of AI for radiology, including an encouraging lung cancer screening study in Glasgow, through to concerns about bias, dissonance, ‘deskilling and never-skilling’, the hidden environment and equity costs, governance, workforce readiness, patient trust – and much more.
You can also watch this playlist. Croakey will be bringing you more in-depth AI health stories from RANZCR’s conference in the coming week, so stay tuned.
Marie McInerney writes:
Dr Curt Langlotz couldn’t be more pleased that the “godfather of AI,” Dr Geoffrey Hinton, recently walked back his prediction that radiology was facing extinction as a profession, because the reality of its AI revolution is far more complex.
Langlotz is Professor of Radiology, Medicine, and Biomedical Data Science and Director of the Center for Artificial Intelligence in Medicine and Imaging (AIMI) at Stanford University.
His opening keynote at #Intelligence26 noted that radiology is leading the way when it comes to AI in medicine – of the 1,500 AI software tools cleared to date by the US Food and Drug Authority, 75 percent are radiology focused.
Langlotz recently led a study showing that in the next five years there will likely be a 33 percent reduction in need for radiologists’ work, but that will come in the context of both “incredible growth” in demand and, in contrast, a flat training rate.
In the video below he talks to Croakey about the leading role of radiology in the AI rollout and the need for augmentation versus replacement of specialist skills. He argues that AI efficiency claims may currently be overblown, but will be important in helping radiologists manage increasing demand, and emphasises the need for patient consent and literacy.
Best AI investments
Governments and health systems should not invest in the “fanciest, newest” AI technology, but rather in mature, shareable data and in getting professions “AI ready” so they can explore its potential and understand its limitations for healthcare.
That was a key message to Australian PM Anthony Albanese from Professor Andre Dekker, a medical physicist and professor of Clinical Data Science at Maastricht University Medical Center and Maastro Clinic in The Netherlands.
Dekker has been instrumental in creating systems that facilitate the Findability, Accessibility, Interoperability, and Reusability (FAIR) principles of health data, including the development of the Personal Health Train (PHT) framework. This is infrastructure that enables data sharing without requiring the information to leave a hospital, thereby addressing ethical and privacy concerns.
He told Croakey at #Intelligence26 that one of the key lessons from radiology and radiation oncology, which are ahead of the curve in AI development, is that “a bit of scepticism is not bad”— particularly in the face of growing pressure from governments and other funders to adopt AI, when it may actually not deliver the promised efficiency gains.
“AI is great, but also if you really want to apply it for the high stakes that we have in healthcare, it really still needs a lot of work,” he said.
In this interview, Dekker also talks about the growing costs of AI, impacts on equity, and privacy and consent.
AI assisted diagnoses
#Intelligence26 did provide some positive news about AI’s potential to support expert diagnosis.
AI supported lung cancer screening research in Glasgow is investigating whether additional tests undertaken alongside a CT scan may help to identify illness much earlier and reduce Emergency Department demand.
Professor David Lowe is Clinical Director of Innovation at the University of Glasgow, a Consultant in Emergency Medicine at the city’s Queen Elizabeth University Hospital, and Clinical Director for Health Innovation for the Scottish Government.
In the United Kingdom, he said, the Emergency Department is the location where 80 percent of people get their heart disease failure diagnosis, 50 percent get a chronic obstructive pulmonary disease (COPD) diagnosis, and 50 percent get their lung cancer diagnosis.
Those numbers are, he said, “incredibly motivating” for clinicians to use AI to develop a pathway to earlier detection.
In the interview below he spoke to Croakey about the GALACTIC-1 study, which is exploring whether structured reporting of the CT scan, combined with additional tests (blood sample, ECG heart trace, and spirometry breathing test), can help identify conditions such as COPD, pulmonary fibrosis, coronary artery disease, heart failure, and early signs of osteoporosis.
Impact on medical education?
With AI now capable of passing medical exams, does it represent an existential threat to medical education?
That big question was posed by Associate Professor David Kok, a radiation oncologist at the Peter MacCallum Cancer Centre and Head of Cancer Science at the University of Melbourne.
Kok’s academic and research activities are focused on using AI in medical education to improve student learning, such as through personalised AI tutoring, while maintaining educational integrity. It’s a challenging field now it is almost impossible to detect the use of AI in some examination environments, like take home tests.
There are already concerns around ‘AI de-skilling’, where clinicians lose skills through not having to use them. However, Kok warned of the perhaps more worrying “never-skilling”, saying there are nascent signs that student reliance on AI for cognitive tasks is leading to errors in assessments and practice.
Education will have to change fundamentally he said, and that will require significant resources, warning that AI tools can be “extraordinarily good prediction machines” but can be both “persuasive and unsafe”.
Sustainability and equity
Dr Florence (Flo) Doo is urging radiologists to grasp the dual nature of AI, which can serve as a tool to enhance sustainability in medical imaging, such as through shorter CT scan times, but will also drive massive growth in greenhouse gas emissions, with the potential to worsen climate and equity.
Doo is an abdominal radiologist, imaging informaticist, and co-lead of the AI-Enabled Medical Imaging team at the Center for Applied AI within the University of Maryland Institute for Health Computing.
In her address on sustainable technology, titled “Is your AI worth its watts”, she detailed the AI data centres’ huge demands for power and water, and the implications for equity of cost and access.
She is concerned also that while people are now more conscious about climate change and may, for example, avoid using disposable plastics when they’re eating out, that personal restraint does not necessarily translate into work environments. This blindspot may be a huge “hidden cost” in an AI-driven information environment.
Unanswered questions
Dr Martin Gunn was chair of RANZCR’s AI Advisory Committee when planning began for #Intelligence26, a follow-up to the College’s first dedicated AI conference in 2018, when the prospect of artificial intelligence was still only “very aspirational”, he recalls.
“Fast forward to 2026, and now we use AI in our everyday practices, and we’ve seen what it’s like, what impact it’s having on our workforce, what impact it’s having on our patients, what impact it’s having on our business,” he said.
Still, there remain many questions to be answered about its value in practice, including a “tension…between promise and reality” and the possible impacts of automation and confirmation bias on clinical decision-making.
Gunn is the chief medical officer and a radiologist at a private practice group in New Zealand, and a subspecialist radiologist in body/abdominal imaging and emergency radiology.
He also talked to Croakey about regulation in Australia and New Zealand, and patient consent issues – particularly his concern that once AI supports in radiology become “the standard of care and practice”, patients who refuse their use may put themselves at a disadvantage in healthcare.
Whose decision-making?
In medicine, where there is an inherent degree of error, “we know our blind spots”, says US clinician and academic Professor Ruth Carlos from Columbia University, a board-certified radiologist specialising in abdominal imaging, and the first woman editor in chief of the Journal of the American College of Radiology (JACR).
But as AI constitutes a new entity with its own opinions, clinicians and researchers will have to consider what they do when there is dissonance between it and human decision making.
As a scientist, she said, she remains agnostic on the value of AI —“I go where the evidence takes me”.
However, she warned of the risks where clinicians become “overly reliant on it because we haven’t developed our own set of expertise” or “when we plan to implement it in clinical practice and don’t develop metrics that tell us whether it is actually working or fit for purpose”.
Who is not at the table?
For Associate Professor Hyun Soo Ko, AI can unlock the untapped potential of imaging to prevent major health events and reduce inequities. It can also fail to live up to its promise outside of the ‘incubation box’.
Ko presented at #Intelligence26 on the potential for ‘opportunistic screening’. This is making the most of millions of scans done for a particular reason, which contain other critical health information “that we’re not really interrogating”.
“With AI, we have the opportunity to leverage all that big data to get more information out of already existing data,” she says.
So long as these investigations are developed, implemented and monitored in sound ethico-legal and practical frameworks, she says they have potential improve general population health outcomes including for often under-served communities, such as Aboriginal and Torres Strait Islander peoples, Māori and Pacific communities, and rural and remote populations.
Ko is a radiologist and clinical researcher at the Peter MacCallum Cancer Centre in Melbourne, and one of the convenors of #Intelligence26.
Rubbish in, rubbish out
Dr Farhannah Aly, another of the #Intelligence26 convenors and current chair of the RANZCR AI Committee, says strong institutional governance is key for the safe and effective use of AI for clinical care — an issue that RANZCR has been working proactively on.
Ahead of the conference, Aly talked about some of the speakers and issues they would be addressing, and about her PhD. It is exploring outcome prediction models for people with cancer, looking particularly at better care for older patients, who are generally an underrepresented group in clinical trials.
Aly told Croakey she is truly excited about the potential of AI for her clinical practice as a radiation oncologist, but it’s “tempered with realism that it’s not so simple just to go from this excitement to implementation in one step. It’s actually a very long process.”
Among her concerns are the quality and use of patient data — “we’re very aware that if you put rubbish into an algorithm, you get rubbish out” — and how to build consumer and patient trust into the way AI is deployed in their clinical care.
• Disclaimer from Marie McInerney: AI assisted with transcribing but all quotes were checked by the journalist.
Bookmark this link to follow Croakey Conference News Service’s coverage of #Intelligence26.
Add Comment