Demographic bias of expert-level vision-language foundation models in medical imaging
Science Advances · 28.03.2025 · 11 (13)
DOI: 10.1126/sciadv.adq0305 ↗Kurz gefasst
Fachartikel in der Zeitschrift Science Advances, erschienen am 28.03.2025. Verfasst von Yuzhe Yang und 8 weiteren Autor:innen. Beteiligte Einrichtung: Department of Electrical Engineering and Computer Science, Massachusetts Institute of Technology, Cambridge, MA, USA. Hauptthemen laut MEDLINE: Artificial Intelligence, Diagnostic Imaging. Die Arbeit wurde bisher 72-mal zitiert (Crossref) und nennt 54 Literaturangaben. Der Volltext ist frei in PubMed Central verfügbar.
Abstract (Originalsprache)
Advances in artificial intelligence (AI) have achieved expert-level performance in medical imaging applications. Notably, self-supervised vision-language foundation models can detect a broad spectrum of pathologies without relying on explicit training annotations. However, it is crucial to ensure that these AI models do not mirror or amplify human biases, disadvantaging historically marginalized groups such as females or Black patients. In this study, we investigate the algorithmic fairness of state-of-the-art vision-language foundation models in chest x-ray diagnosis across five globally sourced datasets. Our findings reveal that compared to board-certified radiologists, these foundation models consistently underdiagnose marginalized groups, with even higher rates seen in intersectional subgroups such as Black female patients. Such biases present over a wide range of pathologies and demographic attributes. Further analysis of the model embedding uncovers its substantial encoding of demographic information. Deploying medical AI systems with biases can intensify preexisting care disparities, posing potential challenges to equitable healthcare access and raising ethical questions about their clinical applications.
Schlagwörter
MeSH-Schlagwörter (MEDLINE)
Autor:innen und Institutionen
- Department of Electrical Engineering and Computer Science, Massachusetts Institute of Technology, Cambridge, MA, USA.
- Department of Electrical and Computer Engineering, University of California, San Diego, La Jolla, CA, USA.
- Paul G. Allen School of Computer Science & Engineering, University of Washington, Seattle, WA, USA.
- Department of Radiology, University of Washington School of Medicine, Seattle, WA, USA.
- OncoRad/Tumor Imaging Metrics Core (TIMC), Department of Radiology, University of Washington, Seattle, WA, USA.
- Department of Electrical and Computer Engineering, University of Washington, Seattle, WA, USA.
ORCID-Profile: Yuzhe Yang, Domenico Mastrodicasa, Wei Wu, Edward J Wang, Shwetak Patel
Zitieren
APA
Yang, Y., Liu, Y., Liu, X., Gulhane, A., Mastrodicasa, D., Wu, W., Wang, E. J., Sahani, D., & Patel, S. (2025). Demographic bias of expert-level vision-language foundation models in medical imaging. Science Advances, 11(13). https://doi.org/10.1126/sciadv.adq0305
Vancouver
Yang Y, Liu Y, Liu X, Gulhane A, Mastrodicasa D, Wu W, et al. Demographic bias of expert-level vision-language foundation models in medical imaging. Science Advances. 2025;11(13). doi:10.1126/sciadv.adq0305
BibTeX
@article{Yang2025,
author = {Yang, Yuzhe and Liu, Yujia and Liu, Xin and Gulhane, Avanti and Mastrodicasa, Domenico and Wu, Wei and Wang, Edward J and Sahani, Dushyant and Patel, Shwetak},
title = {Demographic bias of expert-level vision-language foundation models in medical imaging},
journal = {Science Advances},
year = {2025},
volume = {11},
number = {13},
doi = {10.1126/sciadv.adq0305}
}
EUVIMED