Authors
Yee Seng Ng, Elise Dunning
Published in
Current problems in diagnostic radiology. Nov 06, 2026. Epub Nov 06, 2026.
Abstract
Foundation models represent a major evolution in artificial intelligence, enabling models pretrained on large, diverse datasets to be adapted to a wide range of downstream tasks with substantially less task-specific training data. In radiology, advances in transformer architectures, self-supervised learning, and vision-language models have created new opportunities to integrate image and text understanding within versatile artificial intelligence systems. Potential applications span the radiology workflow and include image interpretation, automated report drafting and summarization, clinical decision support, information retrieval, and development of specialized task-specific models. Among these, AI-assisted report generation represents a particularly promising near-term application, although current performance supports a role as a drafting assistant rather than an autonomous replacement for radiologists. Important barriers to clinical adoption remain, including limitations in model performance, regulatory requirements, and the risk of automation bias. In the near term, foundation models may have their greatest impact by augmenting radiologists and accelerating the development of specialized artificial intelligence applications rather than replacing existing systems or physician expertise. This review describes the technological advances underlying foundation models, highlights their current and emerging applications in radiology, examines key risks and implementation challenges, and considers their potential impact on the future practice of radiology.
PMID:
43035333
Bibliographic data and abstract were imported from PubMed on 16 Nov 2026.
Read full publication at:
Please sign in
to see all details.
Advertisement
Stats
- Recommendations n/a n/a positive of 0 vote(s)
- Views 5
- Comments 0