mlsanityData to 7 Oct 2026

Confounding variables can degrade generalization performance of radiological deep learning models

arXiv 1807.00431 ↗First posted 2 Jul 2018cs.CV cs.LG stat.MLcited by 374

Not on the map. Why?

Abstract (arXiv)

Early results in using convolutional neural networks (CNNs) on x-rays to diagnose disease have been promising, but it has not yet been shown that models trained on x-rays from one hospital or one group of hospitals will work equally well at different hospitals. Before these tools are used for computer-aided diagnosis in real-world clinical settings, we must verify their ability to generalize across a variety of hospital systems. A cross-sectional design was used to train and evaluate pneumonia screening CNNs on 158,323 chest x-rays from NIH (n=112,120 from 30,805 patients), Mount Sinai (42,396 from 12,904 patients), and Indiana (n=3,807 from 3,683 patients). In 3 / 5 natural comparisons, performance on chest x-rays from outside hospitals was significantly lower than on held-out x-rays from the original hospital systems. CNNs were able to detect where an x-ray was acquired (hospital system, hospital department) with extremely high accuracy and calibrate predictions accordingly. The performance of CNNs in diagnosing diseases on x-rays may reflect not only their ability to identify disease-specific imaging findings on x-rays, but also their ability to exploit confounding information. Estimates of CNN performance based on test data from hospital systems used for model training may overstate their likely real-world performance.

Key facts. Cited by 374 papers in the map (data through 7 Oct 2026). First posted on arXiv 2 Jul 2018.

Builds on

No reference list yet.

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2021: 55; 2022: 73; 2023: 56; 2024: 40; 2025: 60; 2026: 90.

374 papers.

Sort:

By area:

  1. 2024Sparse Feature Circuits: Discovering and Editing Interpretable Causal Graphs in Language Modelscited by 412
  2. 2022Last Layer Re-Training is Sufficient for Robustness to Spurious Correlationscited by 379
  3. 2021Interpretable Machine Learning: Fundamental Principles and 10 Grand Challengescited by 364
  4. 2022Domino: Discovering Systematic Errors with Cross-Modal Embeddingscited by 155
  5. 2022Diverse Weight Averaging for Out-of-Distribution Generalizationcited by 148
  6. 2022On Feature Learning in the Presence of Spurious Correlationscited by 145
  7. 2024WARM: On the Benefits of Weight Averaged Reward Modelscited by 145
  8. 2021On Calibration and Out-of-domain Generalizationcited by 134
  9. 2022Goal Misgeneralization: Why Correct Specifications Aren't Enough For Correct Goalscited by 120
  10. 2021Causality-inspired Single-source Domain Generalization for Medical Image Segmentationcited by 118
  11. 2022Model Ratatouille: Recycling Diverse Models for Out-of-Distribution Generalizationcited by 97
  12. 2021MedPerf: Open Benchmarking Platform for Medical Artificial Intelligence using Federated Evaluationcited by 88
38 more
  1. 2021Deep Learning for Chest X-ray Analysis: A Surveycited by 84
  2. 2022Federated Learning Enables Big Data for Rare Cancer Boundary Detectioncited by 83
  3. 2021IAIA-BL: A Case-based Interpretable Deep Learning Model for Classification of Mass Lesions in Digital Mammographycited by 72
  4. 2021On the Impact of Spurious Correlation for Out-of-distribution Detectioncited by 68
  5. 2023Simple and Fast Group Robustness by Automatic Feature Reweightingcited by 68
  6. 2022Probable Domain Generalization via Quantile Risk Minimizationcited by 66
  7. 2022Leakage and the Reproducibility Crisis in ML-based Sciencecited by 63
  8. 2022ProtoPFormer: Concentrating on Prototypical Parts in Vision Transformers for Interpretable Image Recognitioncited by 63
  9. 2021Have We Learned to Explain?: How Interpretability Methods Can Learn to Encode Predictions in their Interpretationscited by 59
  10. 2021Counterfactuals and Causability in Explainable Artificial Intelligence: Theory, Algorithms, and Applicationscited by 59
  11. 2021OpenFL: An open-source framework for Federated Learningcited by 58
  12. 2023Towards Last-layer Retraining for Group Robustness with Fewer Annotationscited by 54
  13. 2021Do Input Gradients Highlight Discriminative Features?cited by 52
  14. 2022Detecting Shortcut Learning for Fair Medical AI using Shortcut Testingcited by 52
  15. 2021Source-Free Adaptation to Measurement Shift via Bottom-Up Feature Restorationcited by 50
  16. 2021The Federated Tumor Segmentation (FeTS) Challengecited by 49
  17. 2022Diagnosing failures of fairness transfer across distribution shift in real-world medical settingscited by 45
  18. 2021Out-of-distribution Generalization in the Presence of Nuisance-Induced Spurious Correlationscited by 41
  19. 2022Spurious Features Everywhere -- Large-Scale Detection of Harmful Spurious Features in ImageNetcited by 41
  20. 2023Fast Diffusion-Based Counterfactuals for Shortcut Removal and Generationcited by 41
  21. 2022Robust and Efficient Medical Imaging with Self-Supervisioncited by 40
  22. 2024The Clever Hans Mirage: A Comprehensive Survey on Spurious Correlations in Machine Learningcited by 39
  23. 2021How I failed machine learning in medical imaging -- shortcomings and recommendationscited by 37
  24. 2023Diagnosing Model Performance Under Distribution Shiftcited by 37
  25. 2022Contrastive Learning for Unsupervised Domain Adaptation of Time Seriescited by 36
  26. 2022Calibrated ensembles can mitigate accuracy tradeoffs under distribution shiftcited by 34
  27. 2025Steering Out-of-Distribution Generalization with Concept Ablation Fine-Tuningcited by 34
  28. 2021An Empirical Framework for Domain Generalization in Clinical Settingscited by 32
  29. 2021QU-BraTS: MICCAI BraTS 2020 Challenge on Quantifying Uncertainty in Brain Tumor Segmentation - Analysis of Ranking Scores and Benchmarking Resultscited by 28
  30. 2022Understanding Machine Learning Practitioners' Data Documentation Perceptions, Needs, Challenges, and Desideratacited by 28
  31. 2021An Information-theoretic Approach to Distribution Shiftscited by 25
  32. 2023Auditing Fairness by Bettingcited by 24
  33. 2021Vision Transformer using Low-level Chest X-ray Feature Corpus for COVID-19 Diagnosis and Severity Quantificationcited by 21
  34. 2023Spuriosity Didn't Kill the Classifier: Using Invariant Predictions to Harness Spurious Featurescited by 21
  35. 2022Eye-gaze-guided Vision Transformer for Rectifying Shortcut Learningcited by 20
  36. 2022Explicit Tradeoffs between Adversarial and Natural Distributional Robustnesscited by 20
  37. 2022First Steps Toward Understanding the Extrapolation of Nonlinear Models to Unseen Domainscited by 20
  38. 2025Steering Language Models with Weight Arithmeticcited by 19

50 most cited of 374 All 374 (JSON)