Economic evaluations of AI applications in radiology: a systematic review
Springer Nature Link, 2026
This systematic review examines an increasingly important question in radiology AI: how do we evaluate whether these tools deliver not only clinical benefit, but also economic value? Reviewing studies published between 2014 and 2025, the authors found 31 economic evaluations of AI applications in radiology, of which only 16 were full economic evaluations. The studies most often focused on direct costs, diagnostic accuracy, and cost-effectiveness, with quality-adjusted life years (QALYs) used as the most common outcome measure.
A key message is that the current evidence base remains highly inconsistent. The included studies varied widely in design, comparators, and metrics, making results difficult to compare across settings. Many evaluations relied on decision-analytic models, especially in opportunistic imaging, and most came from the US and UK, with relatively little evidence from continental Europe. This suggests that the economic case for radiology AI is still being built on a narrow and uneven foundation.
Importantly, the review shows that economic assessments of radiology AI still place too much emphasis on accuracy and direct cost measures, while often overlooking productivity, workflow efficiency, and access to care. These are critical outcomes for real-world implementation, especially in busy radiology services where AI may affect triage, reporting speed, and workforce capacity.
Overall, the paper argues for more robust and standardized economic evaluation methods, supported by international guidance and interdisciplinary collaboration. It is therefore highlighted that responsible AI adoption in radiology will depend not only on technical performance, but also on stronger, more comparable evidence of value in practice.