Harnessing Adversarial Distillation to Customise Debiased, Disease-Specific Pathology Foundation Models for Breast Cancer

2026-08-02

AI Summary

This work addresses the vulnerability of general-purpose foundation models in breast cancer diagnosis to distribution shifts and non-biological biases, which often lead to shortcut learning and neglect of critical subtle morphological features. To mitigate this, the authors propose SmartStu, a novel framework that integrates adversarial distillation, multi-teacher ensembling, and artifact-injected self-supervised auxiliary tasks to guide a lightweight student model in identifying and suppressing confounding spurious features. Evaluated across three external breast cancer cohorts, SmartStu achieves over 30-fold model compression while matching or surpassing the original large model in both balanced accuracy and AUC, demonstrating an efficient, robust, and debiased approach to tailored pathological analysis for breast cancer.

Abstract

Pathology foundation models (PFMs) provide strong tissue representations and have become central to digital pathology. However, deployment in disease-specific settings is limited by 1) the high computational cost of billion-parameter PFMs and 2) distribution mismatch and non-biological bias inherited from pan-cancer, multi-centre pre-training, including site-specific signatures and imbalanced disease prevalence. These factors can encourage shortcut learning and under-emphasise subtle morphology required for reliable modelling of a specific cancer type. We present SmartStu (a Smart Student), a framework to customise compact, breast-cancer-specific PFMs via distillation whilst mitigating confounding. SmartStu distils representations from multiple teacher PFMs into a lightweight student backbone. Crucially, we introduce adversarial distillation that leverages a dedicated noise model trained to predict nuisance, edge-dominated cues on the distillation set. Using this noise model as a counterexample, the adversarial objective encourages the student to recognise, yet suppress, features predictive of nuisance targets. We further incorporate multi-teacher ensemble distillation and an auxiliary self-supervised objective with artefact injection. We validate SmartStu on three external cohorts (Yale HER2, SLN-Breast, and BRACS) with multiple tiny backbones. SmartStu yields breast-cancer-specific PFMs that are over $30\times$ smaller than general PFMs whilst largely preserving, and sometimes improving, downstream performance measured by balanced accuracy (bAcc) and AUC. Code is available at https://github.com/zwchen03/advDistall.

Problem

• pathology foundation models
• distribution mismatch
• non-biological bias
• shortcut learning
• computational cost

Innovation

• adversarial distillation
• pathology foundation models
• bias mitigation
• knowledge distillation
• breast cancer