Agentic AI Systems in Radiology: Design, Applications, Evaluation, and Challenges.
Authors
Affiliations (11)
Affiliations (11)
- Diagnostic and Interventional Radiology, University Hospital Zurich, University of Zurich, Zurich, Switzerland.
- Center for Artificial Intelligence in Medicine and Imaging, Stanford University, 1701 Page Mill Rd, Palo Alto, CA 94304.
- Department of Radiology, Stanford University, Palo Alto, Calif.
- HOPPR, Menlo Park, Calif.
- Diagnostic and Interventional Radiology, University Hospital Aachen, Aachen, Germany.
- Lab for AI in Medicine, University Hospital Aachen, Aachen, Germany.
- Else Kröner Fresenius Center for Digital Health, Faculty of Medicine, University Hospital Carl Gustav Carus, TUD Dresden University of Technology, Dresden, Germany.
- Department of Biosystems Science and Engineering, ETH Zurich, Basel, Switzerland.
- Department of Biomedical Data Science, Stanford University, Palo Alto, Calif.
- Department of Quantitative Biomedicine, University of Zurich, Zurich, Switzerland.
- Institute of Computer Science, Zurich University of Applied Sciences, Zurich, Switzerland.
Abstract
Building agents that perceive and act on their environment with a degree of autonomy has long been a focus of artificial intelligence research. This pursuit has become far more practical with the emergence of large language models (LLMs) capable of using natural language to integrate information, follow instructions, and perform forms of reasoning and planning across a wide range of tasks. With its multimodal data streams and orchestrated workflows spanning multiple systems, radiology is uniquely suited to benefit from agents that can adapt to context and automate complex, repetitive tasks. In radiology, LLMs and their multimodal variants have already demonstrated promising performance for individual tasks such as information extraction and report summarization. However, using LLMs in isolation underutilizes their potential to support complex, multistep workflows where decisions depend on evolving context from multiple information sources. Equipping LLMs with external tools and feedback mechanisms enables them to drive systems that exhibit a spectrum of autonomy, ranging from semiautomated workflows to more adaptive agents capable of managing complex processes. This review article examines the design of LLM-driven agentic systems, explores representative radiology applications, discusses evaluation frameworks, and outlines key challenges, including error cascades, tool-use effectiveness, and integration with health information technology infrastructure.