Stain-Aware Wavelet Regularization for Instant Adversarial Purification in Histopathology

2026-06-07

AI Summary

Deep neural networks in computational pathology are highly vulnerable to adversarial perturbations, particularly because high-frequency adversarial noise closely resembles diagnostically critical fine histological structures, thereby undermining clinical reliability. To address this challenge, this work proposes a real-time adversarial purification framework that introduces, for the first time, stain-aware wavelet-domain regularization. Leveraging the Haar wavelet transform, the method performs multi-scale frequency decoupling and incorporates biological priors derived from hematoxylin and eosin (H&E) staining channels to guide frequency modulation. This approach effectively disentangles adversarial noise from pathology-relevant structures while preserving both textural details and spectral fidelity of tissue samples. Under adversarial attack, the proposed method achieves up to a 10.69% improvement in robustness over baseline approaches.

Abstract

Deep learning has become prevalent in computational pathology pipelines that support tasks such as cancer screening and digital pathology analysis. However, the susceptibility of neural networks to adversarial perturbations raises safety concerns for reliable deployment in clinical practice. In histopathological images, this challenge is exacerbated by the difficulty of distinguishing high-frequency adversarial noise from subtle and diagnostically relevant tissue structures. To address this issue, we propose Stain-Aware Wavelet Regularization (SAWR), an adversarial purification framework that leverages multi-level wavelet-domain regularization based on Haar transform to hierarchically disentangle adversarial perturbations from diagnostic structural information. This spectral constraint is further extended to individual histological channels, enabling stain-specific frequency regulation consistent with the biological properties of Hematoxylin and Eosin. When integrated into an instant purification framework, SAWR improves adversarial robustness by up to 10.69\% over the baseline approach, while maintaining texture and spectral fidelity under adversarial perturbations.

Problem

• adversarial perturbations
• histopathology
• computational pathology
• adversarial robustness
• diagnostic structures

Innovation

• Stain-Aware Wavelet Regularization
• Adversarial Purification
• Histopathology
• Haar Wavelet Transform
• Multi-level Frequency Regulation