Toward Robust Diagnosis: A Contour Attention Preserving Adversarial Defense for COVID-19 Detection

TL;DR

This paper introduces a novel adversarial defense method for COVID-19 diagnosis from CT images, leveraging contour attention preservation to improve robustness and generalization across different datasets.

Contribution

The proposed Contour Attention Preserving (CAP) method enhances adversarial robustness in COVID-19 CT diagnosis by integrating lung cavity edge features into the attention mechanism.

Findings

Achieves state-of-the-art adversarial defense performance

Demonstrates improved generalization across datasets

Effectively preserves lung cavity contours during diagnosis

Abstract

As the COVID-19 pandemic puts pressure on healthcare systems worldwide, the computed tomography image based AI diagnostic system has become a sustainable solution for early diagnosis. However, the model-wise vulnerability under adversarial perturbation hinders its deployment in practical situation. The existing adversarial training strategies are difficult to generalized into medical imaging field challenged by complex medical texture features. To overcome this challenge, we propose a Contour Attention Preserving (CAP) method based on lung cavity edge extraction. The contour prior features are injected to attention layer via a parameter regularization and we optimize the robust empirical risk with hybrid distance metric. We then introduce a new cross-nation CT scan dataset to evaluate the generalization capability of the adversarial robustness under distribution shift. Experimental results indicate that the proposed method achieves state-of-the-art performance in multiple adversarial defense and generalization tasks. The code and dataset are available at https://github.com/Quinn777/CAP.