{"id":329085,"date":"2025-12-05T09:04:13","date_gmt":"2025-12-05T09:04:13","guid":{"rendered":"https:\/\/www.newsbeep.com\/au\/329085\/"},"modified":"2025-12-05T09:04:13","modified_gmt":"2025-12-05T09:04:13","slug":"why-are-there-still-so-many-radiologists","status":"publish","type":"post","link":"https:\/\/www.newsbeep.com\/au\/329085\/","title":{"rendered":"Why are there still so many radiologists?"},"content":{"rendered":"

This article is an on-site version of The AI Shift newsletter. Premium subscribers can sign up here<\/a> to get the newsletter delivered every Thursday. Standard subscribers can upgrade to Premium here<\/a>, or explore<\/a> all FT newsletters<\/p>\n

Hello, and welcome back to The AI Shift, our newsletter about AI and the world of work.<\/p>\n

This week we ask: why are there still so many radiologists? Back in 2016, Geoffrey Hinton, often described as the \u201cgodfather of AI\u201d, said<\/a> they were already on borrowed time. \u201cI think if you work as a radiologist, you\u2019re like the coyote that\u2019s already over the edge of the cliff but hasn\u2019t yet looked down,\u201d he told his audience. \u201cPeople should stop training radiologists now. It\u2019s just completely obvious that within five years deep learning is going to do better than radiologists\u2009.\u2009.\u2009.\u2009It might be ten years, but we\u2019ve got plenty of radiologists already.\u201d<\/p>\n

So how has that prediction turned out, and what can we learn from it? Over to John for a look at the data.<\/p>\n

John writes<\/p>\n

I think it\u2019s useful here to tease apart the different predictions that were explicit or implicit in Hinton\u2019s remarks and look at how each stacks up.<\/p>\n

There are two main claims: that AI would be outperforming radiologists by the early to mid 2020s, and that as a result demand for radiologists was roughly at its peak in 2016.<\/p>\n

If we start with the first, and we take Hinton as referring to deep learning models\u2019 capacity to detect and interpret tell-tale patterns in X-rays and other medical images, then you could argue that for some conditions he was proved right within a year, let alone five or ten. As early as 2017, CheXNet \u2014 a specialist algorithm trained to detect pneumonia from chest X-rays \u2014 was shown to outperform practising radiologists<\/a> with decades of experience. Since then such tools have rapidly matured from experimental to regulator-approved and are now used by healthcare providers in many high-income countries, as chronicled in this excellent deep-dive<\/a> by Deena Mousa in Works in Progress magazine. Last year the UK\u2019s National Health Service adopted an AI diagnostic tool<\/a> for use across its radiology departments after a 2023 trial<\/a> helped to detect lung cancers at an earlier stage and cut the time from initial scan to treatment.<\/p>\n

But there\u2019s a catch, which brings us on to Hinton\u2019s second prediction. While he didn\u2019t directly claim radiologist employment was about to decline, he did imply that these advances in pattern detection and interpretation meant demand for radiologists\u2019 skills and services had peaked. This does not appear to have been the case. In his native UK, the total number of radiologists employed by the NHS has climbed by more than 40 per cent<\/a> since 2016. In Canada, where he is currently based, the number of new radiologists taking up residencies has been trending upwards and reached a record high this year<\/a>. And in the US, where we have so far seen the strongest evidence of AI job displacement<\/a> in other sectors, last year\u2019s 1,378 new radiologist recruits represented a 20 per cent increase<\/a> since Hinton\u2019s remarks, while pay for the specialty has grown more rapidly<\/a> than most others.<\/p>\n

That\u2019s because almost all of the AI tools in use by healthcare providers today are being used by radiologists, not instead of them. The tools keep getting better, and now match or outperform experienced radiologists even after factoring in false positives or negatives<\/a>, but the fact that both human and AI remain fallible means it makes far more sense to pair them up than for one to replace the other. Two pairs of eyes can come to a quicker and more accurate judgment, one spotting or correcting something the other missed. And in high-stakes settings where the costs of a mistake can be astronomical, the downside risk from an error by a fully autonomous AI radiologist is huge.<\/p>\n

Sarah writes<\/p>\n

In order to expand on what you\u2019ve found, John, I spoke to a radiologist about what AI in her profession looks like on the ground. Amaka Offiah, who is a consultant paediatric radiologist and a professor in paediatric musculoskeletal imaging at the University of Sheffield in the UK, reinforced what your reporting shows. She told me radiologists in the UK now use AI for all sorts of tasks, from helping to detect and measure lung nodules in CT scans which can be signs of cancer, to AI-enabled MRI scanners which can cut the time of a scan, so you can \u201cget that patient on the table and off the table quicker.\u201d A survey last year by the Royal College of Radiologists (where Offiah is vice-president for clinical radiology) found that 69 per cent of radiology departments were using AI in clinical practice, up from 54 per cent in 2023.<\/p>\n

But the survey<\/a> also found something striking: only 6 per cent of clinical directors said AI tools had reduced their workload, 37 per cent reported an increase in workload, and the rest reported no change. When I asked Offiah what explained those results (which are, after all, the precise opposite of what Hinton predicted) she gave me a number of reasons.\u00a0<\/p>\n

Some were the sorts of teething issues that one might expect to get better over time, such as trouble integrating AI with existing IT infrastructure. Others were more fundamental. AI tools create new tasks and responsibilities, such as \u201cpost-deployment monitoring\u201d, which involves \u201cauditing to make sure [the tool] is still performing at the level of accuracy [that was] on the tin,\u201d as she put it. In addition, by speeding up some parts of the workflow, AI can lead to more work and bottlenecks elsewhere (those with good memories will recall we found the same issue in software development in a previous edition<\/a> of this newsletter). If AI speeds up the rate at which you can do MRI scans, for example, it means more images for radiologists to report on.<\/p>\n

As for reading the images, although there are many AI tools now available and more being trialled<\/a> for certain tasks, there aren\u2019t plentiful training datasets for every type of patient and every type of problem. A model trained on adult X-rays, for example, won\u2019t be reliable for children. Offiah, a paediatric radiologist, says getting robust data on children is hard, both because of ethical approvals and consent from parents, and also because they change so much as they grow. \u201cEverybody thinks children are small adults, but they\u2019re not \u2014 it\u2019s quite a different physiology,\u201d she said. \u201cI\u2019m diagnosing X-rays of an 8 week old foetus, right up to 17 year olds.\u201d<\/p>\n

The other difference between AI and human radiologists, she said, is that the latter will notice things they\u2019re not necessarily looking for. \u201cLots of AI is trained for simple and single tasks \u2014 it might be looking at the nodule in the lung, but it hasn\u2019t necessarily been trained to recognise the bone metastasis in the humerus, which you can see on the same CT scan.\u201d <\/p>\n

And finally, of course, radiologists do a lot more than just interpret images. They also decide what imaging is needed, and they combine the images with their judgment, experience, and knowledge of the patient\u2019s history to diagnose and monitor the progress of treatment, usually via discussion with other doctors in multidisciplinary teams.<\/p>\n

Put all that together with the context of an ageing population and growing demand for imaging of all kinds, and you can see why Offiah and the Royal College of Radiologists are concerned about a shortage of radiologists<\/a>, not their displacement.<\/p>\n

\u201cAI will assist radiologists, but will not replace them,\u201d she said, in a prediction of her own. \u201cI could even dare to say: will never replace them.\u201d<\/p>\n

So what have we learned?<\/p>\n

I find this a fascinating demonstration of why even if AI really can do some of the most high-value parts of someone\u2019s job, it doesn\u2019t mean displacement (even of those few tasks let alone the job as a whole) is inevitable. Though I also can\u2019t help noticing a parallel to driverless cars, which were simply too risky to ever go fully autonomous until they weren\u2019t<\/a>.<\/p>\n

Sarah<\/p>\n

I don\u2019t want to be too unfair to Hinton, who is hardly the only person to have made big predictions in this space, but I think the story of radiologists should be a reminder to technologists not to make sweeping assertions about the future of professions they don\u2019t intimately understand. If we had indeed stopped training radiologists in 2016, we\u2019d be in a real mess today.<\/p>\n

Recommended reading<\/p>\n

A fascinating podcast<\/a> episode on how AI is changing warfare (Sarah)<\/p>\n

I want to re-up Deena Mousa\u2019s thorough interrogation of the AI radiologist question<\/a> for Works in Progress, which digs into some of the issues I touched on here in more detail (John)<\/p>\n

Recommended newsletters for you<\/p>\n

The Lex Newsletter \u2014 Lex, our investment column, breaks down the week\u2019s key themes, with analysis by award-winning writers. Sign up here<\/a><\/p>\n

Working It \u2014 Everything you need to get ahead at work, in your inbox every Wednesday. Sign up here<\/a><\/p>\n","protected":false},"excerpt":{"rendered":"This article is an on-site version of The AI Shift newsletter. Premium subscribers can sign up here to…\n","protected":false},"author":2,"featured_media":329086,"comment_status":"","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[34],"tags":[64,63,137,500],"class_list":{"0":"post-329085","1":"post","2":"type-post","3":"status-publish","4":"format-standard","5":"has-post-thumbnail","7":"category-healthcare","8":"tag-au","9":"tag-australia","10":"tag-health","11":"tag-healthcare"},"_links":{"self":[{"href":"https:\/\/www.newsbeep.com\/au\/wp-json\/wp\/v2\/posts\/329085","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.newsbeep.com\/au\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.newsbeep.com\/au\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/au\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/au\/wp-json\/wp\/v2\/comments?post=329085"}],"version-history":[{"count":0,"href":"https:\/\/www.newsbeep.com\/au\/wp-json\/wp\/v2\/posts\/329085\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/au\/wp-json\/wp\/v2\/media\/329086"}],"wp:attachment":[{"href":"https:\/\/www.newsbeep.com\/au\/wp-json\/wp\/v2\/media?parent=329085"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.newsbeep.com\/au\/wp-json\/wp\/v2\/categories?post=329085"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.newsbeep.com\/au\/wp-json\/wp\/v2\/tags?post=329085"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}