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
FullText Text:
  Availability: 0
CustomLinks:
  – Url: https://dx.doi.org/doi:10.1007/s11604-026-02021-6
    Name: EDS - Springer Nature Journals (s7799221)
    Category: fullText
    Text: View record at Springer
  – Url: https://resolver.ebsco.com/c/fiv2js/result?sid=EBSCO:cmedm&genre=article&issn=1867108X&ISBN=&volume=44&issue=10&date=20261001&spage=1863&pages=1863-1873&title=Japanese journal of radiology&atitle=Assessing%20and%20mitigating%20demographic%20bias%20in%20large%20language%20models%20for%20diagnostic%20radiology.&aulast=Horiuchi%20D&id=DOI:10.1007/s11604-026-02021-6
    Name: Full Text Finder (for New FTF UI) (ns324271)
    Category: fullText
    Text: Full Text Finder
    MouseOverText: Full Text Finder
Header DbId: cmedm
DbLabel: MEDLINE
An: 42247082
AccessLevel: 3
PubType: Academic Journal
PubTypeId: academicJournal
PreciseRelevancyScore: 0
IllustrationInfo
Items – Name: Title
  Label: Title
  Group: Ti
  Data: Assessing and mitigating demographic bias in large language models for diagnostic radiology.
– Name: Author
  Label: Authors
  Group: Au
  Data: <searchLink fieldCode="AU" term="%22Horiuchi+D%22">Horiuchi D</searchLink>; Department of Diagnostic and Interventional Radiology, Graduate School of Medicine, Osaka Metropolitan University, 1-4-3 Asahi-machi, Abeno-ku, Osaka, 545-8585, Japan.<br /><searchLink fieldCode="AU" term="%22Tatekawa+H%22">Tatekawa H</searchLink>; Department of Diagnostic and Interventional Radiology, Graduate School of Medicine, Osaka Metropolitan University, 1-4-3 Asahi-machi, Abeno-ku, Osaka, 545-8585, Japan.<br /><searchLink fieldCode="AU" term="%22Takita+H%22">Takita H</searchLink>; Department of Diagnostic and Interventional Radiology, Graduate School of Medicine, Osaka Metropolitan University, 1-4-3 Asahi-machi, Abeno-ku, Osaka, 545-8585, Japan.<br /><searchLink fieldCode="AU" term="%22Mitsuyama+Y%22">Mitsuyama Y</searchLink>; Department of Artificial Intelligence, Graduate School of Medicine, Osaka Metropolitan University, 1-4-3 Asahi-machi, Abeno-ku, Osaka, 545-8585, Japan.<br /><searchLink fieldCode="AU" term="%22Walston+SL%22">Walston SL</searchLink>; Department of Artificial Intelligence, Graduate School of Medicine, Osaka Metropolitan University, 1-4-3 Asahi-machi, Abeno-ku, Osaka, 545-8585, Japan.<br /><searchLink fieldCode="AU" term="%22Saito+K%22">Saito K</searchLink>; Department of Artificial Intelligence, Graduate School of Medicine, Osaka Metropolitan University, 1-4-3 Asahi-machi, Abeno-ku, Osaka, 545-8585, Japan.<br /><searchLink fieldCode="AU" term="%22Shimono+T%22">Shimono T</searchLink>; Department of Diagnostic and Interventional Radiology, Graduate School of Medicine, Osaka Metropolitan University, 1-4-3 Asahi-machi, Abeno-ku, Osaka, 545-8585, Japan.<br /><searchLink fieldCode="AU" term="%22Miki+Y%22">Miki Y</searchLink>; Department of Diagnostic and Interventional Radiology, Graduate School of Medicine, Osaka Metropolitan University, 1-4-3 Asahi-machi, Abeno-ku, Osaka, 545-8585, Japan.<br /><searchLink fieldCode="AU" term="%22Ueda+D%22">Ueda D</searchLink>; 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.
– Name: TitleSource
  Label: Source
  Group: Src
  Data: <searchLink fieldCode="JN" term="%22101490689%22">Japanese journal of radiology</searchLink> [Jpn J Radiol] 2026 Oct; Vol. 44 (10), pp. 1863-1873. <i>Date of Electronic Publication: </i>2026 Jun 05.
– Name: TypePub
  Label: Publication Type
  Group: TypPub
  Data: Journal Article
– Name: Language
  Label: Language
  Group: Lang
  Data: English
– Name: TitleSource
  Label: Journal Info
  Group: Src
  Data: <i>Publisher: </i><searchLink fieldCode="PB" term="%22Springer%22">Springer </searchLink><i>Country of Publication: </i>Japan <i>NLM ID: </i>101490689 <i>Publication Model: </i>Print-Electronic <i>Cited Medium: </i>Internet <i>ISSN: </i>1867-108X (Electronic) <i>Linking ISSN: </i><searchLink fieldCode="IS" term="%2218671071%22">18671071 </searchLink><i>NLM ISO Abbreviation: </i>Jpn J Radiol <i>Subsets: </i>MEDLINE
– Name: PublisherInfo
  Label: Imprint Name(s)
  Group: PubInfo
  Data: <i>Original Publication</i>: Tokyo : Springer
– Name: SubjectMESH
  Label: MeSH Terms
  Group: Su
  Data: <searchLink fieldCode="MM" term="%22Radiology%22">Radiology*</searchLink>/<searchLink fieldCode="MM" term="%22Radiology+methods%22">methods</searchLink> <br /><searchLink fieldCode="MM" term="%22Large+Language+Models%22">Large Language Models*</searchLink> <br /><searchLink fieldCode="MM" term="%22Diagnostic+Imaging%22">Diagnostic Imaging*</searchLink><br /><searchLink fieldCode="MH" term="%22Humans%22">Humans</searchLink> ; <searchLink fieldCode="MH" term="%22Female%22">Female</searchLink> ; <searchLink fieldCode="MH" term="%22Male%22">Male</searchLink> ; <searchLink fieldCode="MH" term="%22Retrospective+Studies%22">Retrospective Studies</searchLink> ; <searchLink fieldCode="MH" term="%22Middle+Aged%22">Middle Aged</searchLink>
– Name: Abstract
  Label: Abstract
  Group: Ab
  Data: 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.<br />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.<br />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.<br />Conclusion: Large language models exhibited racial and sex biases in diagnostic radiology, and chain-of-thought prompting may help mitigate these biases.<br /> (© 2026. The Author(s).)
– Name: Abstract
  Label: Competing Interests
  Group: Ab
  Data: 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.
– Name: Ref
  Label: References
  Group: RefInfo
  Data: Radiology. 2025 May;315(2):e241674. (PMID: <searchLink fieldCode="PM" term="%2240392092%22">40392092)</searchLink><br />NPJ Digit Med. 2024 Oct 21;7(1):295. (PMID: <searchLink fieldCode="PM" term="%2239433945%22">39433945)</searchLink><br />Radiographics. 2025 Apr;45(4):e240073. (PMID: <searchLink fieldCode="PM" term="%2240048389%22">40048389)</searchLink><br />Radiology. 2023 Jul;308(1):e231040. (PMID: <searchLink fieldCode="PM" term="%2237462501%22">37462501)</searchLink><br />Nat Med. 2025 Jun;31(6):1873-1881. (PMID: <searchLink fieldCode="PM" term="%2240195448%22">40195448)</searchLink><br />Nat Med. 2024 Dec;30(12):3590-3600. (PMID: <searchLink fieldCode="PM" term="%2239313595%22">39313595)</searchLink><br />Lancet Digit Health. 2024 Jan;6(1):e12-e22. (PMID: <searchLink fieldCode="PM" term="%2238123252%22">38123252)</searchLink><br />Radiology. 2024 Jan;310(1):e232756. (PMID: <searchLink fieldCode="PM" term="%2238226883%22">38226883)</searchLink><br />NPJ Digit Med. 2025 Jun 4;8(1):332. (PMID: <searchLink fieldCode="PM" term="%2240467886%22">40467886)</searchLink><br />Radiology. 2024 Jul;312(1):e240273. (PMID: <searchLink fieldCode="PM" term="%2238980179%22">38980179)</searchLink><br />Radiology. 2024 Sep;312(3):e240153. (PMID: <searchLink fieldCode="PM" term="%2239225605%22">39225605)</searchLink><br />Radiology. 2025 Sep;316(3):e241516. (PMID: <searchLink fieldCode="PM" term="%2240985835%22">40985835)</searchLink><br />Eur Radiol. 2025 Jan;35(1):506-516. (PMID: <searchLink fieldCode="PM" term="%2238995378%22">38995378)</searchLink><br />Nat Med. 2023 Aug;29(8):1930-1940. (PMID: <searchLink fieldCode="PM" term="%2237460753%22">37460753)</searchLink><br />Radiology. 2025 Apr;315(1):e240528. (PMID: <searchLink fieldCode="PM" term="%2240298602%22">40298602)</searchLink><br />Clin Imaging. 2024 May;109:110113. (PMID: <searchLink fieldCode="PM" term="%2238552383%22">38552383)</searchLink><br />Jpn J Radiol. 2026 Apr;44(4):627-640. (PMID: <searchLink fieldCode="PM" term="%2241236701%22">41236701)</searchLink><br />JAMA. 2021 Aug 17;326(7):621-627. (PMID: <searchLink fieldCode="PM" term="%2234402850%22">34402850)</searchLink><br />Radiology. 2025 Jan;314(1):e240689. (PMID: <searchLink fieldCode="PM" term="%2239835982%22">39835982)</searchLink><br />Radiol Artif Intell. 2024 Jan;6(1):e230103. (PMID: <searchLink fieldCode="PM" term="%2238294325%22">38294325)</searchLink><br />JAMA Netw Open. 2025 Sep 2;8(9):e2532692. (PMID: <searchLink fieldCode="PM" term="%2240991290%22">40991290)</searchLink><br />Radiographics. 2024 May;44(5):e230067. (PMID: <searchLink fieldCode="PM" term="%2238635456%22">38635456)</searchLink><br />Biometrics. 2000 Jun;56(2):345-51. (PMID: <searchLink fieldCode="PM" term="%2210877288%22">10877288)</searchLink><br />Jpn J Radiol. 2024 Jan;42(1):3-15. (PMID: <searchLink fieldCode="PM" term="%2237540463%22">37540463)</searchLink><br />Radiology. 2023 Sep;308(3):e232082. (PMID: <searchLink fieldCode="PM" term="%2237724974%22">37724974)</searchLink><br />JAMA. 2016 Nov 8;316(18):1863-1864. (PMID: <searchLink fieldCode="PM" term="%2227802482%22">27802482)</searchLink><br />NPJ Digit Med. 2023 Oct 20;6(1):195. (PMID: <searchLink fieldCode="PM" term="%2237864012%22">37864012)</searchLink><br />Radiology. 2024 May;311(2):e232715. (PMID: <searchLink fieldCode="PM" term="%2238771184%22">38771184)</searchLink><br />Nat Med. 2025 Oct;31(10):3283-3289. (PMID: <searchLink fieldCode="PM" term="%2240954311%22">40954311)</searchLink><br />NPJ Digit Med. 2025 May 4;8(1):246. (PMID: <searchLink fieldCode="PM" term="%2240319154%22">40319154)</searchLink>
– Name: GrantInfo
  Label: Grant Information
  Group: Grant
  Data: JPMJPR2521 Precursory Research for Embryonic Science and Technology
– Name: SubjectMinor
  Label: Contributed Indexing
  Group:
  Data: <i>Keywords: </i>Artificial intelligence; Bias; Large language models; Radiology
– Name: DateEntry
  Label: Entry Date(s)
  Group: Date
  Data: <i>Date Created: </i>20260605 <i>Date Completed: </i>20260929 <i>Latest Revision: </i>20261001
– Name: DateUpdate
  Label: Update Code
  Group: Date
  Data: 20261001
– Name: DOI
  Label: DOI
  Group: ID
  Data: 10.1007/s11604-026-02021-6
– Name: AN
  Label: PMID
  Group: ID
  Data: 42247082
PLink https://search.ebscohost.com/login.aspx?direct=true&site=eds-live&db=cmedm&AN=42247082
RecordInfo BibRecord:
  BibEntity:
    Identifiers:
      – Type: doi
        Value: 10.1007/s11604-026-02021-6
    Languages:
      – Code: eng
        Text: English
    PhysicalDescription:
      Pagination:
        StartPage: 1863
    Subjects:
      – SubjectFull: Humans
        Type: general
      – SubjectFull: Female
        Type: general
      – SubjectFull: Male
        Type: general
      – SubjectFull: Retrospective Studies
        Type: general
      – SubjectFull: Middle Aged
        Type: general
      – SubjectFull: Radiology methods
        Type: general
      – SubjectFull: Large Language Models
        Type: general
      – SubjectFull: Diagnostic Imaging
        Type: general
    Titles:
      – TitleFull: Assessing and mitigating demographic bias in large language models for diagnostic radiology.
        Type: main
  BibRelationships:
    HasContributorRelationships:
      – PersonEntity:
          Name:
            NameFull: Horiuchi D
      – PersonEntity:
          Name:
            NameFull: Tatekawa H
      – PersonEntity:
          Name:
            NameFull: Takita H
      – PersonEntity:
          Name:
            NameFull: Mitsuyama Y
      – PersonEntity:
          Name:
            NameFull: Walston SL
      – PersonEntity:
          Name:
            NameFull: Saito K
      – PersonEntity:
          Name:
            NameFull: Shimono T
      – PersonEntity:
          Name:
            NameFull: Miki Y
      – PersonEntity:
          Name:
            NameFull: Ueda D
    IsPartOfRelationships:
      – BibEntity:
          Dates:
            – D: 01
              M: 10
              Text: 2026 Oct
              Type: published
              Y: 2026
          Identifiers:
            – Type: issn-electronic
              Value: 1867-108X
          Numbering:
            – Type: volume
              Value: 44
            – Type: issue
              Value: 10
          Titles:
            – TitleFull: Japanese journal of radiology
              Type: main
ResultId 1