Adversarial Attacks on Medical Hyperspectral Imaging Exploiting Spectral-Spatial Dependencies and Multiscale Features
Yunrui Gu, Zhenzhe Gao, Cong Kong, Zhaoxia Yin
TL;DR
This paper addresses the vulnerability of medical hyperspectral imaging (MHSI) classifiers to adversarial perturbations by identifying two key fragilities: reliance on local pixel dependencies for tissue structure and dependence on multiscale spectral-spatial representations for hierarchical features. It introduces two targeted attacks, Local Pixel Dependency Attack and Multiscale Information Attack, and a joint adversarial framework that combines them to degrade lesion-region classification while keeping perturbations imperceptible. Through experiments on brain and MDC datasets, the approach achieves substantial erosion of tumor/cancer accuracy across models and defenses, revealing vulnerabilities that are not captured by global metrics. The findings underscore the need for robust, structure-aware defenses in clinical MHSI applications and motivate future work incorporating domain priors to improve lesion-region robustness.
Abstract
Medical hyperspectral imaging (HSI) enables accurate disease diagnosis by capturing rich spectral-spatial tissue information, but recent advances in deep learning have exposed its vulnerability to adversarial attacks. In this work, we identify two fundamental causes of this fragility: the reliance on local pixel dependencies for preserving tissue structure and the dependence on multiscale spectral-spatial representations for hierarchical feature encoding. Building on these insights, we propose a targeted adversarial attack framework for medical HSI, consisting of a Local Pixel Dependency Attack that exploits spatial correlations among neighboring pixels, and a Multiscale Information Attack that perturbs features across hierarchical spectral-spatial scales. Experiments on the Brain and MDC datasets demonstrate that our attacks significantly degrade classification performance, especially in tumor regions, while remaining visually imperceptible. Compared with existing methods, our approach reveals the unique vulnerabilities of medical HSI models and underscores the need for robust, structure-aware defenses in clinical applications.