Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2023 Jun;307(5):e231259.
doi: 10.1148/radiol.231259.

Evaluating GPT4 on Impressions Generation in Radiology Reports

Zhaoyi Sun #  1 , Hanley Ong #  2 , Patrick Kennedy  2 , Liyan Tang  3 , Shirley Chen  2 , Jonathan Elias  4 , Eugene Lucas  5 , George Shih  2 , Yifan Peng  1
Affiliations

Evaluating GPT4 on Impressions Generation in Radiology Reports

Zhaoyi Sun et al. Radiology. 2023 Jun.
No abstract available

PubMed Disclaimer

Conflict of interest statement

Z.S. No relevant relationships. H.O. No relevant relationships. P.K. No relevant relationships. L.T. No relevant relationships. S.C. No relevant relationships. J.E. No relevant relationships. E.L. No relevant relationships. G.S. No relevant relationships. Y.P. Various federal grants (2145640, NSF CAREER Award, NSF 4R00LM013001, NIH/NLM.

Figures

Figure 1:
Figure 1:
Questionnaire for human evaluation.
Figure 2:
Figure 2:
Performance of GPT-4 in Impression generation in human evaluations. (A) Coherence refers to the ability of the Impressions to build a coherent body of information about a topic through sentence-to-sentence connections. (B) Comprehensiveness evaluates whether the Impressions contain sufficient information to convey the abnormal Findings. (C) Factual consistency measures whether the Findings support the impressions. (D) Harmfulness refers to the potential of Impressions to lead to physical or psychologic harm.
See this image and copyright information in PMC

Comment in

References

    1. Hartung MP, Bickle IC, Gaillard F, Kanne JP. How to Create a Great Radiology Report. RadioGraphics 2020;40(6):1658–1670. - PubMed
    1. Gundogdu B, Pamuksuz U, Chung JH, et al. Customized Impression Prediction from Radiology Reports Using BERT and LSTMs. IEEE Transactions on Artificial Intelligence. 2021; 1–1.
    1. GPT-4. OpenAI. https://openAI.com/gpt-4. Accessed May 25, 2023.
    1. Tang L, Sun Z, Idnay B, et al. Evaluating Large Language Models on Medical Evidence Summarization. medRxiv 2023.04.22.23288967. Posted April 24, 2023. Accessed May 15, 2023. - PMC - PubMed
    1. Bhayana R, Krishna S, Bleakney RR. Performance of ChatGPT on a Radiology Board-style Examination: Insights into Current Strengths and Limitations. Radiology 2023;307(5):e230582. - PubMed

LinkOut - more resources