Physics-Grounded Adversarial Stain Augmentation with Calibrated Coverage Guarantees

TL;DR

This paper introduces CASA, a stain augmentation method with calibrated coverage guarantees, improving histopathology model robustness across hospitals by adversarially perturbing stain parameters within a statistically calibrated budget.

Contribution

CASA is a novel stain augmentation technique that uses a principled, statistically calibrated adversarial approach in stain parameter space, providing coverage guarantees for unseen centers.

Findings

CASA achieves 93.9% slide-level accuracy on Camelyon17-WILDS, outperforming existing methods.

CASA provides the highest worst-group accuracy among compared methods.

CASA's adversarial augmentation improves model robustness across multiple centers.

Abstract

Stain variation across hospitals degrades histopathology models at deployment. Existing augmentation methods perturb color spaces with arbitrary hyperparameters, lacking both a principled budget and coverage guarantees for unseen centers. We propose \textbf{C}alibrated \textbf{A}dversarial \textbf{S}tain \textbf{A}ugmentation (\textbf{CASA}), which performs adversarial augmentation in the Macenko stain parameter space with a budget calibrated from multi-center statistics via the DKW inequality. On Camelyon17-WILDS (5 seeds), CASA achieves $93.9\% \pm 1.6\%$ slide-level accuracy -- outperforming HED-strong ($88.4\% \pm 7.3\%$), RandStainNA ($85.2\% \pm 6.7\%$), and ERM ($63.9\% \pm 11.3\%$) -- with the highest worst-group accuracy ($84.9\% \pm 0.9\%$) among all 10 compared methods.