Marginal Contribution of Current AI Application in Radiology in the Western World

Huisman et al. highlight the context-dependent value of AI in radiology. In their article, “The Emperor Has Few Clothes: A Realistic Appraisal of Current AI in Radiology,” they present a critical perspective on AI’s actual impact in the Western setting. Merel Huisman, a radiologist and clinical epidemiologist with a special focus on AI, sees it as her mission to caution the community against being swept away by AI’s shininess. After all, history in fields like pharmaceuticals teaches us to be wary of some very expensive innovations in the light of healthcare accessibility.

“In our opinion, by adding narrow AI tools on top of current workflows, no upfront costs are saved since radiologists still must conduct full reads and maintain oversight,” the authors postulate. They also consider the fear of AI replacing radiologists outdated.

They argue that AI’s true value in radiology will only emerge when it can autonomously manage simpler tasks, reducing workload without compromising safety.

The authors raise concerns over “complexity overhead,” where technological advancements inadvertently increase costs and environmental demands, noting the unknown yet potentially substantial carbon footprint of medical AI. By maintaining a focus solely on diagnostic tasks, they see no big future for the technique in radiology.

“Use-cases that would come in very handy for a radiologist but are not getting nearly as much attention as diagnostic tools are automated protocolling, personalized hanging protocols, and auto-summary of findings.”

Their call to action urges the community to critically evaluate AI claims, prioritizing the safe automation of tasks that support healthcare accessibility rather than focusing on marginal quality enhancements.

Article: The emperor has few clothes: a realistic appraisal of current AI in radiology

Authors: Merel Huisman, Bram van Ginneken & Hugh Harvey