• Economic Value of AI in Radiology: A Systematic Review

    Isabel Molwitz, Inka Ristow, Jennifer Erley, Tugba Akinci D'Antonoli, Ali S Tejani, Michail E Klontzas, Merel Huisman, Gerhard Adam, Stephan Nüesch, Lisa Adams
    Radiol Artif Intell. 2025 Nov 19:e250090. doi: 10.1148/ryai.250090. Online ahead of print.

    Abstract

    Purpose To summarize the evidence of artificial intelligence's (AI) economic value across the radiologic workflow. Materials and Methods A comprehensive search of PubMed, Business Source Ultimate, and EconLit was conducted for original research articles published between January 2010 and November 2024. Medical Subject Headings and keywords included "artificial intelligence/machine learning/deep learning/natural language processing," "radiology," and "economic value/cost/budget/revenue/efficiency." Studies were selected based on explicit quantification of economic outcomes, excluding those with only soft outcome criteria like time savings without cost quantification. Study quality was assessed using the Criteria for Health Economic Quality Evaluation (CHEQUE). Results From the initial 1,879 search results, 21 studies (1%) met the inclusion criteria. The majority evaluated machine learning tools (10/21[48%], nine on deep learning), followed by computer-assisted diagnostics (CAD; 7/21[33%]), natural language processing (NLP; 2/21[10%]), and hypothetical AI models (2/21[10%]). AI demonstrated economic value through cost savings or incremental cost-effectiveness ratios in resource-intensive tasks, when accuracy matched human performance and costs were fixed. For instance, AI-based lung cancer screening achieved incremental cost savings of up to USD 242 per patient. AI increased costs when specificity was below humans' or when using pay-per-use models; as observed with CAD systems raising mammography screening costs by up to USD 19 per patient. In fast tasks such as radiograph evaluations, AI showed value in settings with radiologist shortages. AI reduced costs through protocol optimization and increased revenue via improved follow-up compliance. Conclusion AI's value in radiology is context-dependent, varying with task complexity, examination volume, and implementation model. Further high-quality economic evaluations are essential.