Exploring the Clinical Translation of Generative Models Like ChatGPT: Promise and Pitfalls in Radiology, From Patients to Population Health

Florence X Doo; Tessa S Cook; Eliot L Siegel; Anupam Joshi; Vishwa Parekh; Ameena Elahi; Paul H Yi

doi:10.1016/j.jacr.2023.07.007

Exploring the Clinical Translation of Generative Models Like ChatGPT: Promise and Pitfalls in Radiology, From Patients to Population Health

J Am Coll Radiol. 2023 Sep;20(9):877-885. doi: 10.1016/j.jacr.2023.07.007. Epub 2023 Jul 17.

Authors

Florence X Doo¹, Tessa S Cook², Eliot L Siegel³, Anupam Joshi⁴, Vishwa Parekh⁵, Ameena Elahi⁶, Paul H Yi⁷

Affiliations

¹ Director of Innovation, University of Maryland Medical Intelligent Imaging Center (UM2ii), Baltimore, Maryland; Member, Committee on Economics in Academic Radiology, under the ACR Commission on Economics. Electronic address: [email protected].
² Vice Chair for Practice Transformation, Department of Radiology, Perelman School of Medicine at the University of Pennsylvania, Philadelphia, Pennsylvania; Fellowship Director, Imaging Informatics, and Chief, 3-D and Advanced Imaging, Department of Radiology, Penn Medicine, Philadelphia, Pennsylvania; Chair, Society for Imaging Informatics in Medicine; and Vice Chair, ACR Commission on Patient- and Family-Centered Care; Chair, RAHSR Affinity Group. Electronic address: https://twitter.com/asset25.
³ Vice Chair, Research Information Systems, University of Maryland, Baltimore, Maryland; Lead, Radiology and Nuclear Medicine Diagnostics, US Department of Veterans Affairs Veterans Integrated Services Network; Chief, Imaging, US Department of Veterans Affairs Maryland Healthcare System; Radiology AI Senior Consultant. Electronic address: https://twitter.com/EliotSiegel.
⁴ Oros Family Professor and Chair, Computer Science and Electrical Engineering, University of Maryland, Baltimore, Maryland; Director, University of Maryland, Baltimore County, Center for Cybersecurity; Director, CyberScholars Program; Associate Editor, IEEE Transactions on Dependable and Secure Computing.
⁵ Technical Director, University of Maryland Medical Intelligent Imaging (UM2ii) Center, Baltimore, Maryland; Review Editor, Frontiers in Oncology. Electronic address: https://twitter.com/vishwa_parekh.
⁶ University of Pennsylvania, Philadelphia, Pennsylvania; Application Manager, Information Services, Penn Medicine, Philadelphia, Pennsylvania; Informatics Operations Director, RAD-AID International. Electronic address: https://twitter.com/AmeenaElahi.
⁷ Director, University of Maryland Medical Intelligent Imaging (UM2ii) Center, Baltimore, Maryland; Vice Chair, Society of Imaging Informatics in Medicine Program Planning Committee; Associate Editor, Radiology: Artificial Intelligence. Electronic address: https://twitter.com/PaulYiMD.

PMID: 37467871
DOI: 10.1016/j.jacr.2023.07.007

Abstract

Generative artificial intelligence (AI) tools such as GPT-4, and the chatbot interface ChatGPT, show promise for a variety of applications in radiology and health care. However, like other AI tools, ChatGPT has limitations and potential pitfalls that must be considered before adopting it for teaching, clinical practice, and beyond. We summarize five major emerging use cases for ChatGPT and generative AI in radiology across the levels of increasing data complexity, along with pitfalls associated with each. As the use of AI in health care continues to grow, it is crucial for radiologists (and all physicians) to stay informed and ensure the safe translation of these new technologies.

Keywords: ChatGPT; generative artificial intelligence; large language models; limitations; radiology.

MeSH terms

Artificial Intelligence
Humans
Population Health*
Radiography
Radiologists
Radiology*