Hierarchical Self-Supervised Adversarial Training for Robust Vision Models in Histopathology

2025-03-13

AI Summary

To address insufficient adversarial robustness in histopathological image analysis, this paper proposes Hierarchical Self-supervised Adversarial Training (HSAT), the first framework to incorporate patient–slide–patch hierarchical structural modeling into self-supervised adversarial training. HSAT constructs semantically consistent adversarial examples via multi-level contrastive learning and integrates domain-specific pathological priors to achieve multi-granularity feature alignment. On the OpenSRH dataset, HSAT improves robustness by 54.31% under white-box attacks and incurs only a 3–4% performance drop under black-box attacks—substantially outperforming baseline methods (which suffer 25–30% degradation) and establishing a new state-of-the-art benchmark. Key contributions are: (i) the first exploitation of intrinsic hierarchical structure in histopathology to guide adversarial example generation; and (ii) the deep integration of hierarchical modeling with self-supervised contrastive learning and PGD-based adversarial training, jointly optimizing both discriminability and robustness.

Technology Category

• Computer Vision: Adversarial Attacks & Robustness
• Machine Learning: Adversarial Learning & Robustness
• Search and Optimization: Adversarial Search

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems
• Graph Algorithms and Modeling for the Web: Algorithms and analysis for heterogeneous, signed, attributed, multi-relational, temporal, higher-order, and annotated Web-related graphs

Abstract

Adversarial attacks pose significant challenges for vision models in critical fields like healthcare, where reliability is essential. Although adversarial training has been well studied in natural images, its application to biomedical and microscopy data remains limited. Existing self-supervised adversarial training methods overlook the hierarchical structure of histopathology images, where patient-slide-patch relationships provide valuable discriminative signals. To address this, we propose Hierarchical Self-Supervised Adversarial Training (HSAT), which exploits these properties to craft adversarial examples using multi-level contrastive learning and integrate it into adversarial training for enhanced robustness. We evaluate HSAT on multiclass histopathology dataset OpenSRH and the results show that HSAT outperforms existing methods from both biomedical and natural image domains. HSAT enhances robustness, achieving an average gain of 54.31% in the white-box setting and reducing performance drops to 3-4% in the black-box setting, compared to 25-30% for the baseline. These results set a new benchmark for adversarial training in this domain, paving the way for more robust models. Our Code for training and evaluation is available at https://github.com/HashmatShadab/HSAT.

Problem

• Address adversarial attacks in histopathology vision models
• Exploit hierarchical structure for robust adversarial training
• Improve model robustness in white-box and black-box settings

Innovation

• Hierarchical Self-Supervised Adversarial Training (HSAT)
• Multi-level contrastive learning for adversarial examples
• Enhanced robustness in histopathology vision models