MeDi: Metadata-Guided Diffusion Models for Mitigating Biases in Tumor Classification

2025-06-20

AI Summary

Deep learning models for tumor classification in histopathological images suffer from poor generalization to underrepresented subpopulations due to biases introduced by staining protocols, scanning devices, hospitals, and demographic variations—leading to shortcut learning and prediction disparities. To address this, we propose a metadata-guided diffusion generative framework that explicitly incorporates clinical metadata (e.g., stain type, institution, demographic attributes) into the conditional diffusion model architecture—enabling zero-shot, high-fidelity synthesis of histopathological images across diverse subpopulations. Leveraging TCGA pretraining and fine-grained metadata alignment, our method generates high-quality images for unseen subpopulations to debias downstream classifiers. Experiments demonstrate that classifiers trained on our synthesized data achieve an average accuracy gain of 8.3% and a 62% reduction in Equalized Odds difference on subpopulation-shifted test sets—significantly outperforming conventional data augmentation and robust training baselines.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Ethics, Bias, and Fairness
• Philosophy and Ethics of AI: Bias, Fairness & Equity

Application Category

• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets

Abstract

Deep learning models have made significant advances in histological prediction tasks in recent years. However, for adaptation in clinical practice, their lack of robustness to varying conditions such as staining, scanner, hospital, and demographics is still a limiting factor: if trained on overrepresented subpopulations, models regularly struggle with less frequent patterns, leading to shortcut learning and biased predictions. Large-scale foundation models have not fully eliminated this issue. Therefore, we propose a novel approach explicitly modeling such metadata into a Metadata-guided generative Diffusion model framework (MeDi). MeDi allows for a targeted augmentation of underrepresented subpopulations with synthetic data, which balances limited training data and mitigates biases in downstream models. We experimentally show that MeDi generates high-quality histopathology images for unseen subpopulations in TCGA, boosts the overall fidelity of the generated images, and enables improvements in performance for downstream classifiers on datasets with subpopulation shifts. Our work is a proof-of-concept towards better mitigating data biases with generative models.

Problem

• Mitigating biases in tumor classification models
• Addressing robustness to varying clinical conditions
• Balancing training data for underrepresented subpopulations

Innovation

• Metadata-guided diffusion model for bias mitigation
• Targeted augmentation of underrepresented subpopulations
• High-quality synthetic histopathology image generation