MIMM-X: Disentangling Spurious Correlations for Medical Image Analysis

2025/11/28 by Fay, Louisa, Reguigui, Hajer, Yang, Bin +2
Computer Science · #Generative Adversarial Networks and Image Synthesis #Explainable Artificial Intelligence (XAI) #AI in cancer detection

paper · doi:10.48550/arxiv.2511.22990

Abstract

Deep learning models can excel on medical tasks, yet often experience spurious correlations, known as shortcut learning, leading to poor generalization in new environments. Particularly in medical imaging, where multiple spurious correlations can coexist, misclassifications can have severe consequences. We propose MIMM-X, a framework that disentangles causal features from multiple spurious correlations by minimizing their mutual information. It enables predictions based on true underlying causal relationships rather than dataset-specific shortcuts. We evaluate MIMM-X on three datasets (UK Biobank, NAKO, CheXpert) across two imaging modalities (MRI and X-ray). Results demonstrate that MIMM-X effectively mitigates shortcut learning of multiple spurious correlations.

Citations

• Benchmarking Dependence Measures to Prevent Shortcut Learning in Medical Imaging
• CheXpert Plus: Augmenting a Large Chest X-ray Dataset with Text Radiology Reports, Patient Demographics and Additional Image Formats
• Debiasing Counterfactuals In the Presence of Spurious Correlations
• Right for the Wrong Reason: Can Interpretable ML Techniques Detect Spurious Correlations?
• Improving Out-of-Distribution Robustness via Selective Augmentation
• Counterfactual Invariance to Spurious Correlations: Why and How to Pass Stress Tests
• On the limits of cross-domain generalization in automated X-ray prediction
• Adversarial Domain Adaptation with Domain Mixup
• Distributionally Robust Neural Networks for Group Shifts: On the Importance of Regularization for Worst-Case Generalization
• CheXpert: A Large Chest Radiograph Dataset with Uncertainty Labels and\n Expert Comparison
• Deep Over-sampling Framework for Classifying Imbalanced Data
• Densely Connected Convolutional Networks
• Unsupervised Domain Adaptation by Backpropagation
• The German National Cohort: aims, study design and organization
• Measuring and testing dependence by correlation of distances

Cited by

• Did Models Sufficient Learn? Attribution-Guided Training via Subset-Selected Counterfactual Augmentation

Related

• Benchmarking Dependence Measures to Prevent Shortcut Learning in Medical Imaging
• Debiasing Counterfactuals In the Presence of Spurious Correlations
• Preventing Shortcut Learning in Medical Image Analysis through Intermediate Layer Knowledge Distillation from Specialist Teachers
• When More is Less: Incorporating Additional Datasets Can Hurt Performance By Introducing Spurious Correlations
• Unlearning Spurious Correlations in Chest X-ray Classification