We always update Medicross medical news as well as notable medical news around the world, read the latest news about our center.
AI in Radiology: Why Radiologists Are More Needed Than Ever
Will artificial intelligence replace the radiologist? It is one of the most-asked questions in medicine today.
In a recent online talk hosted by Morton & Partners, AI in Radiology: Assist Us or Replace Us?, body radiologist Dr Seymour van der Merwe gave a clear answer. AI is changing how radiologists work, but not whether they are needed. Used well, it makes them faster, more focused and more valuable to their patients and referring doctors.
A prediction that didn’t come true
In 2016, AI pioneer Geoffrey Hinton famously said that hospitals should stop training radiologists, because deep learning would outperform them within five years. Dr van der Merwe was starting his registrarship at the time.
Ten years on, the picture looks very different. Radiology was expected to be among AI’s first casualties, yet demand for radiologists keeps growing, and Hinton himself has reportedly softened his view.
“Instead of shrinking, the demand for human radiologists is exploding.”
Dr van der Merwe explained this through the Jevons paradox: when a resource becomes more efficient to use, demand for it rises. When imaging becomes faster and more accessible, more scans are requested, and every one still needs expert interpretation.
A tool that gives time back
The strongest case for AI in radiology is efficiency. Across the studies Dr van der Merwe reviewed, AI consistently took on time-consuming, repetitive work:
- In Sweden’s large MASAI breast-screening trial, AI support reduced radiologists’ screen-reading workload by 44% while detecting more cancers.
- In lung imaging, AI cut reporting time from around three minutes to thirty seconds per case.
- AI triage of breast screening shortened the time to biopsy by around 30%.
The point is not that the machine replaces the reader. It handles counting, measuring and flagging, so the radiologist can concentrate on what matters most.
For Dr van der Merwe, AI belongs alongside earlier innovations such as the move from film to digital PACS: a tool that changed the workflow without replacing the clinician.
How Morton & Partners uses AI
At Morton & Partners, every AI platform is used as a second set of eyes, never on its own.
“The AI diagnosis never goes out unless it’s been reviewed by a doctor.”
The radiologist reads the study, reviews what the AI has flagged and makes the final clinical decision. The platforms in use include:
- Gleamer: detects fractures and bone lesions on X-rays, and automatically measures joint angles in the ankle, hip and elsewhere.
- Quantib: supports prostate MRI by highlighting areas of concern and producing a 3D view of where lesions sit, helping urologists and their patients plan biopsy or surgery.
- Lunit: supports breast imaging by assessing breast density and flagging suspicious areas on mammograms.
- Aidoc: prioritises urgent findings on emergency CT, such as pulmonary embolism, brain haemorrhage and large-vessel occlusion.
Many of Morton’s MRI and CT scanners also have AI built in, adjusting settings automatically to shorten scan times and reduce radiation dose.
Why the radiologist remains essential
Dr van der Merwe was candid about what AI cannot do, and why those limits make the radiologist more important, not less.
AI can’t see the whole patient, so it can’t explain itself. Algorithms work from the image alone and reach conclusions in ways even their developers can’t fully trace. A tool may flag a lung nodule without weighing the patient’s history, symptoms or blood results. That clinical context is exactly what the radiologist brings, and it is what allows a diagnosis to be explained to patients and referring doctors.
“How would you feel if you went to a doctor and the response was: I don’t know why you’re ill, but my computer says you’re ill?”
AI can be wrong, and it isn’t trained on our patients. Algorithms still miss findings, and tools tuned for high sensitivity often over-call, so a steady stream of false alarms can lead to flags being ignored. Most models are also built on data from Europe and the US. In South Africa, where disease patterns, equipment and clinical workflows differ, expert judgement and local experience are essential to separate real findings from false alarms.
Patients want a doctor. Research on patient attitudes is consistent: patients accept AI as a support tool, but not as a replacement for the radiologist. They are uncomfortable with AI acting on its own, especially when it comes to making or delivering a diagnosis, and they prefer a model where doctor and technology work together.
One study of 212 participants found that 91% wanted a radiologist to review any AI diagnosis, and 79% were comfortable with AI offering a second opinion after the radiologist had made theirs. Only 15% would accept AI as the sole decision-maker, even when its accuracy was very high.
What patients value most is the human element: oversight, empathy and clear communication. They want a safe space to ask questions and to discuss their results with someone who can explain them with sensitivity. Many worry that relying too heavily on AI could make care feel impersonal, reduce shared decision-making and weaken the trust between patient and doctor. The message is clear: patients still place a human expert above AI alone.
Even as patients increasingly upload their own scans to public AI tools, Dr van der Merwe sees the radiologist’s role growing:
“That’s where the human radiologist will play a role: to bring all of this together.”
The real question
The evening’s title asked whether AI will assist radiologists or replace them. Dr van der Merwe suggested that this may be the wrong question. AI is not a competitor. It is becoming an essential part of the toolkit, and the radiologists who embrace it will be best placed to serve their patients.
“Radiologists who use AI will replace radiologists who don’t.”
Dr Seymour van der Merwe is a body radiologist at Morton & Partners Radiologists. He presented “AI in Radiology: Assist Us or Replace Us?” online on 5 November 2026.