Improving Adversarial Robustness of ECG Classification Based on Lipschitz Constraints and Channel Activation Suppression

Abstract

Deep neural networks (DNNs) are increasingly important in the medical diagnosis of electrocardiogram (ECG) signals. However, research has shown that DNNs are highly vulnerable to adversarial examples, which can be created by carefully crafted perturbations. This vulnerability can lead to potential medical accidents. This poses new challenges for the application of DNNs in the medical diagnosis of ECG signals. This paper proposes a novel network Channel Activation Suppression with Lipschitz Constraints Net (CASLCNet), which employs the Channel-wise Activation Suppressing (CAS) strategy to dynamically adjust the contribution of different channels to the class prediction and uses the 1-Lipschitz’s ℓ∞ distance network as a robust classifier to reduce the impact of adversarial perturbations on the model itself in order to increase the adversarial robustness of the model. The experimental results demonstrate that CASLCNet achieves ACCrobust scores of 91.03% and 83.01% when subjected to PGD attacks on the MIT-BIH and CPSC2018 datasets, respectively, which proves that the proposed method in this paper enhances the model’s adversarial robustness while maintaining a high accuracy rate.

Convolutional Neural Network-Based ECG Signal Classification Model: A Study on Addressing Class Imbalance and Enhancing Model Interpretability

“…In the task of ECG signal classification, deep learning models, especially CNNs, have demonstrated their high efficiency [1][2][3]18]. However, the decision-making processes of these models often lack transparency, widely referred to as the "black box" problem [6,25,26].…”

“…The precise automatic classification of Electrocardiogram (ECG) signals is of paramount importance for the early diagnosis and treatment of cardiac conditions [1]. Despite certain advancements in ECG signal classification achieved by traditional machine learning methods, the complexity of ECG data necessitates the development of more refined models to capture and identify key signals of cardiac abnormalities [2,3].…”

“…Compared to the existing research, the innovation of this study is mainly reflected in the following two aspects: (1). Application of UMAP Technology: This study innovatively applies UMAP technology to the field of ECG signal classification to visualize the high-dimensional semantic features extracted by the middle layer of CNNs.…”