Academic Journal
Προβολή σε EDS

Assessing and mitigating demographic bias in large language models for diagnostic radiology.

Λεπτομέρειες βιβλιογραφικής εγγραφής
Τίτλος: Assessing and mitigating demographic bias in large language models for diagnostic radiology.
Συγγραφείς: Horiuchi D; Department of Diagnostic and Interventional Radiology, Graduate School of Medicine, Osaka Metropolitan University, 1-4-3 Asahi-machi, Abeno-ku, Osaka, 545-8585, Japan., Tatekawa H; Department of Diagnostic and Interventional Radiology, Graduate School of Medicine, Osaka Metropolitan University, 1-4-3 Asahi-machi, Abeno-ku, Osaka, 545-8585, Japan., Takita H; Department of Diagnostic and Interventional Radiology, Graduate School of Medicine, Osaka Metropolitan University, 1-4-3 Asahi-machi, Abeno-ku, Osaka, 545-8585, Japan., Mitsuyama Y; Department of Artificial Intelligence, Graduate School of Medicine, Osaka Metropolitan University, 1-4-3 Asahi-machi, Abeno-ku, Osaka, 545-8585, Japan., Walston SL; Department of Artificial Intelligence, Graduate School of Medicine, Osaka Metropolitan University, 1-4-3 Asahi-machi, Abeno-ku, Osaka, 545-8585, Japan., Saito K; Department of Artificial Intelligence, Graduate School of Medicine, Osaka Metropolitan University, 1-4-3 Asahi-machi, Abeno-ku, Osaka, 545-8585, Japan., Shimono T; Department of Diagnostic and Interventional Radiology, Graduate School of Medicine, Osaka Metropolitan University, 1-4-3 Asahi-machi, Abeno-ku, Osaka, 545-8585, Japan., Miki Y; Department of Diagnostic and Interventional Radiology, Graduate School of Medicine, Osaka Metropolitan University, 1-4-3 Asahi-machi, Abeno-ku, Osaka, 545-8585, Japan., Ueda D; Department of Artificial Intelligence, Graduate School of Medicine, Osaka Metropolitan University, 1-4-3 Asahi-machi, Abeno-ku, Osaka, 545-8585, Japan. ai.labo.ocu@gmail.com.; Center for Health Science Innovation, Osaka Metropolitan University, 1-4-3 Asahi-machi, Abeno-ku, Osaka, 545-8585, Japan. ai.labo.ocu@gmail.com.
Πηγή: Japanese journal of radiology [Jpn J Radiol] 2026 Oct; Vol. 44 (10), pp. 1863-1873. Date of Electronic Publication: 2026 Jun 05.
Τύπος έκδοσης: Journal Article
Γλώσσα: English
Στοιχεία περιοδικού: Publisher: Springer Country of Publication: Japan NLM ID: 101490689 Publication Model: Print-Electronic Cited Medium: Internet ISSN: 1867-108X (Electronic) Linking ISSN: 18671071 NLM ISO Abbreviation: Jpn J Radiol Subsets: MEDLINE
Imprint Name(s): Original Publication: Tokyo : Springer
Ιατρικοί όροι (MeSH): Radiology*/methods , Large Language Models* , Diagnostic Imaging*, Humans ; Female ; Male ; Retrospective Studies ; Middle Aged
Περίληψη: Purpose: Large language models (LLMs) are increasingly integrated into radiology workflows, but their demographic biases have not been evaluated in diagnostic radiology. This study aimed to investigate racial and sex biases in the diagnostic performance of LLMs (text-only and vision models) in radiology and to evaluate whether prompting strategies mitigate these biases.
Materials and Methods: This retrospective study included consecutive Diagnosis Please cases published in Radiology from April 1998 to October 2024, excluding cases with sex-specific diseases. For each case, eight race-sex scenarios were generated by altering four race/ethnicity categories (Asian, Black, Hispanic, White) and two sex categories (male, female). Three LLMs (GPT-5, Claude Sonnet 4.5, Gemini 2.5 Flash) were evaluated as text-only models (medical history and imaging findings) and vision models (medical history and images) using three prompting strategies (basic, self-consistency, chain-of-thought prompting). Generalized estimating equations were used to compare diagnostic accuracy across race/ethnicity, sex, and prompting strategies.
Results: A total of 286 cases were included. Across models and conditions, ten significant race-related and four significant sex-related differences in diagnostic accuracy were observed. Among the four race/ethnicity groups, Black patients were most likely to have significantly lower accuracy (four of the ten statistically significant race-related comparisons [40%]) and least likely to have significantly higher accuracy (1/10 [10%]). Vision models with female patients under basic prompting showed a higher number of significant race-related differences (6/10 [60%]) than vision models with male patients and all text-only models. Text-only and vision models showed no statistically significant differences in diagnostic accuracy across prompting strategies (p = 0.78 and 0.95, respectively); basic and self-consistency prompting produced ten and four significant race- or sex-related differences, respectively, whereas no significant differences were observed with chain-of-thought prompting.
Conclusion: Large language models exhibited racial and sex biases in diagnostic radiology, and chain-of-thought prompting may help mitigate these biases.
(© 2026. The Author(s).)
Competing Interests: Declarations. Competing interest: The authors of this manuscript declare no relationships with any companies, whose products or services may be related to the subject matter of the article. Ethics approval: Institutional Review Board approval was not required because this study used only published cases. Informed consent: Written informed consent was not required for this study because this study used only published cases.
References: Radiology. 2025 May;315(2):e241674. (PMID: 40392092)
NPJ Digit Med. 2024 Oct 21;7(1):295. (PMID: 39433945)
Radiographics. 2025 Apr;45(4):e240073. (PMID: 40048389)
Radiology. 2023 Jul;308(1):e231040. (PMID: 37462501)
Nat Med. 2025 Jun;31(6):1873-1881. (PMID: 40195448)
Nat Med. 2024 Dec;30(12):3590-3600. (PMID: 39313595)
Lancet Digit Health. 2024 Jan;6(1):e12-e22. (PMID: 38123252)
Radiology. 2024 Jan;310(1):e232756. (PMID: 38226883)
NPJ Digit Med. 2025 Jun 4;8(1):332. (PMID: 40467886)
Radiology. 2024 Jul;312(1):e240273. (PMID: 38980179)
Radiology. 2024 Sep;312(3):e240153. (PMID: 39225605)
Radiology. 2025 Sep;316(3):e241516. (PMID: 40985835)
Eur Radiol. 2025 Jan;35(1):506-516. (PMID: 38995378)
Nat Med. 2023 Aug;29(8):1930-1940. (PMID: 37460753)
Radiology. 2025 Apr;315(1):e240528. (PMID: 40298602)
Clin Imaging. 2024 May;109:110113. (PMID: 38552383)
Jpn J Radiol. 2026 Apr;44(4):627-640. (PMID: 41236701)
JAMA. 2021 Aug 17;326(7):621-627. (PMID: 34402850)
Radiology. 2025 Jan;314(1):e240689. (PMID: 39835982)
Radiol Artif Intell. 2024 Jan;6(1):e230103. (PMID: 38294325)
JAMA Netw Open. 2025 Sep 2;8(9):e2532692. (PMID: 40991290)
Radiographics. 2024 May;44(5):e230067. (PMID: 38635456)
Biometrics. 2000 Jun;56(2):345-51. (PMID: 10877288)
Jpn J Radiol. 2024 Jan;42(1):3-15. (PMID: 37540463)
Radiology. 2023 Sep;308(3):e232082. (PMID: 37724974)
JAMA. 2016 Nov 8;316(18):1863-1864. (PMID: 27802482)
NPJ Digit Med. 2023 Oct 20;6(1):195. (PMID: 37864012)
Radiology. 2024 May;311(2):e232715. (PMID: 38771184)
Nat Med. 2025 Oct;31(10):3283-3289. (PMID: 40954311)
NPJ Digit Med. 2025 May 4;8(1):246. (PMID: 40319154)
Grant Information: JPMJPR2521 Precursory Research for Embryonic Science and Technology
Contributed Indexing: Keywords: Artificial intelligence; Bias; Large language models; Radiology
Entry Date(s): Date Created: 20260605 Date Completed: 20260929 Latest Revision: 20261001
Update Code: 20261001
DOI: 10.1007/s11604-026-02021-6
PMID: 42247082
Βάση Δεδομένων: MEDLINE
Περιγραφή
ISSN:1867-108X
DOI:10.1007/s11604-026-02021-6