mlsanityData to 7 Oct 2026

Risk of Bias in Chest Radiography Deep Learning Foundation Models

arXiv 2209.02965 ↗First posted 7 Sep 2022cs.LG cs.AI cs.CVcited by 26

Not on the map. Why?

Abstract (arXiv)

Purpose: To analyze a recently published chest radiography foundation model for the presence of biases that could lead to subgroup performance disparities across biological sex and race. Materials and Methods: This retrospective study used 127,118 chest radiographs from 42,884 patients (mean age, 63 [SD] 17 years; 23,623 male, 19,261 female) from the CheXpert dataset collected between October 2002 and July 2017. To determine the presence of bias in features generated by a chest radiography foundation model and baseline deep learning model, dimensionality reduction methods together with two-sample Kolmogorov-Smirnov tests were used to detect distribution shifts across sex and race. A comprehensive disease detection performance analysis was then performed to associate any biases in the features to specific disparities in classification performance across patient subgroups. Results: Ten out of twelve pairwise comparisons across biological sex and race showed statistically significant differences in the studied foundation model, compared with four significant tests in the baseline model. Significant differences were found between male and female (P < .001) and Asian and Black patients (P < .001) in the feature projections that primarily capture disease. Compared with average model performance across all subgroups, classification performance on the 'no finding' label dropped between 6.8% and 7.8% for female patients, and performance in detecting 'pleural effusion' dropped between 10.7% and 11.6% for Black patients. Conclusion: The studied chest radiography foundation model demonstrated racial and sex-related bias leading to disparate performance across patient subgroups and may be unsafe for clinical applications.

Key facts. Cited by 26 papers in the map (data through 7 Oct 2026). Builds on 13 of its 35 references that are in the map. First posted on arXiv 7 Sep 2022.

Builds on

13 of 35 references.

References · 13

  1. 2016Densely Connected Convolutional Networkscited by 6,691
  2. 2021On the Opportunities and Risks of Foundation Modelscited by 3,961
  3. 2019CheXpert: A Large Chest Radiograph Dataset with Uncertainty Labels and Expert Comparisoncited by 1,410
  4. 2019Hidden Stratification Causes Clinically Meaningful Failures in Machine Learning for Medical Imagingcited by 215
  5. 2019Causality matters in medical imagingcited by 193
  6. 2021Reading Race: AI Recognises Patient's Racial Identity In Medical Imagescited by 155
  7. 2020CheXclusion: Fairness gaps in deep chest X-ray classifierscited by 153
  8. 2021Deep Learning for Chest X-ray Analysis: A Surveycited by 84
  9. 2020On the limits of cross-domain generalization in automated X-ray predictioncited by 61
  10. 2022Robust and Efficient Medical Imaging with Self-Supervisioncited by 40
  11. 2022Self-supervised Learning from 100 Million Medical Imagescited by 31
  12. 2022Potential sources of dataset bias complicate investigation of underdiagnosis by machine learning algorithmscited by 19
1 more
  1. 2022Debiasing Deep Chest X-Ray Classifiers using Intra- and Post-processing Methodscited by 15

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2023: 1; 2024: 9; 2025: 8; 2026: 8.

26 papers.

Sort:

By area:

  1. 2024Foundation Models in Radiology: What, How, When, Why and Why Notcited by 26
  2. 2024Demographic Bias of Expert-Level Vision-Language Foundation Models in Medical Imagingcited by 24
  3. 2024EVA-X: A Foundation Model for General Chest X-ray Analysis with Self-supervised Learningcited by 22
  4. 2024Slicing Through Bias: Explaining Performance Gaps in Medical Image Analysis using Slice Discovery Methodscited by 14
  5. 2025Foundation Models in Medical Imaging: A Review and Outlookcited by 7
  6. 2024Low-Rank Knowledge Decomposition for Medical Foundation Modelscited by 5
  7. 2024LoRKD: Low-Rank Knowledge Decomposition for Medical Foundation Modelscited by 4
  8. 2025Preventing Shortcut Learning in Medical Image Analysis through Intermediate Layer Knowledge Distillation from Specialist Teacherscited by 4
  9. 2025Fair Foundation Models for Medical Image Analysis: Challenges and Perspectivescited by 3
  10. 2025Bias and Generalizability of Foundation Models across Datasets in Breast Mammographycited by 3
  11. 2025Generalist Models in Medical Image Segmentation: A Survey and Performance Comparison with Task-Specific Approachescited by 2
  12. 2023Post-hoc Orthogonalization for Mitigation of Protected Feature Bias in CXR Embeddingscited by 1
14 more
  1. 2025Understanding Dataset Bias in Medical Imaging: A Case Study on Chest X-rayscited by 1
  2. 2026DUCX: Decomposing Unfairness in Tool-Using Chest X-ray Agentscited by 1
  3. 2024The Importance of Model Inspection for Better Understanding Performance Characteristics of Graph Neural Networkscited by 0
  4. 2024Generalizing Orthogonalization for Models with Non-Linearitiescited by 0
  5. 2024Does Data-Efficient Generalization Exacerbate Bias in Foundation Models?cited by 0
  6. 2025Predict Patient Self-reported Race from Skin Histological Imagescited by 0
  7. 2025Multi Anatomy X-Ray Foundation Modelcited by 0
  8. 2026The Impact of Preprocessing Methods on Racial Encoding and Model Robustness in CXR Diagnosiscited by 0
  9. 2026ViX-Ray: A Vietnamese Chest X-Ray Dataset for Vision-Language Modelscited by 0
  10. 2026Intersectional Disentangling of Temporal and Acquisition Bias in Fetal Ultrasoundcited by 0
  11. 2026On-Device Continual Learning with Dual-Stage Buffer and Dynamic Loss for Point-of-Care Pneumonia Diagnosiscited by 0
  12. 2026Effect of Demographic Bias on Skin Lesion Classificationcited by 0
  13. 2026Benchmarking the Robustness of Foundation Models for Mammography under Domain Shiftcited by 0
  14. 2026Subgroup performance analysis of adaptation strategies for chest X-ray foundation modelscited by 0