Demographic bias of expert-level vision-language foundation models in medical imaging
Top Cited Papers
Open Access
- 28 March 2025
- journal article
- Published by American Association for the Advancement of Science (AAAS) in Science Advances
- Vol. 11 (13) , eadq0305
Abstract
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.Keywords
All Related Versions
This publication has 43 references indexed in Scilit:
- Security of medical images for telemedicine: a systematic reviewMultimedia Tools and Applications, 2022
- Underdiagnosis bias of artificial intelligence algorithms applied to chest radiographs in under-served patient populationsNature Medicine, 2021
- The reduction of race and gender bias in clinical treatment recommendations using clinician peer networks in an experimental settingNature Communications, 2021
- PadChest: A large chest x-ray image dataset with multi-label annotated reportsMedical Image Analysis, 2020
- The influence of patient race on the use of diagnostic imaging in United States emergency departments: data from the National Hospital Ambulatory Medical Care surveyBMC Health Services Research, 2020
- Racial/Ethnic Disparities in Follow-Up Adherence for Incidental Pulmonary Nodules: An Application of a Cascade-of-Care FrameworkJournal of the American College of Radiology, 2020
- Ethical limitations of algorithmic fairness solutions in health care machine learningThe Lancet Digital Health, 2020
- International evaluation of an AI system for breast cancer screeningNature, 2020
- Human–machine partnership with artificial intelligence for chest radiograph diagnosisnpj Digital Medicine, 2019
- Estimation of the Youden Index and its Associated Cutoff PointBiometrical Journal, 2005