The Real Cost of AI in Radiology

Artificial intelligence is rapidly transforming radiology, yet its financial impact remains surprisingly opaque. In a systematic review of nearly 2,000 articles, researchers found that only 21 studies—barely 1 per cent—provided explicit economic data alongside their clinical findings.

The results reveal a nuanced picture where one size does not fit all. AI proves most valuable in resource-intensive tasks; for example, lung cancer screening programmes achieved savings of up to USD 242 per patient when the software matched human accuracy and operated on fixed costs. In these scenarios, AI effectively streamlined protocol optimisation and improved patient follow-up compliance.

However, the technology is not a universal money-spinner. Costs actually rose in some scenarios, particularly when AI specificity fell below human standards or when hospitals utilised pay-per-use models. Computer-assisted diagnostics for mammography, for instance, increased costs by up to USD 19 per patient. Ultimately, the review concludes that AI’s economic worth is highly context-dependent, relying heavily on task complexity and the specific implementation model chosen.