Enhance CNN Robustness Against Noises for Classification of 12-Lead ECG with Variable Length (2020-08-08T00:00:00.000000Z)

TL;DR

A CNN is designed for classification of 12-lead ECG signals with variable length, and three defense methods are applied to improve robustness of this CNN against adversarial noises and white noises, with a minimal reduction in accuracy on clean data.

Abstract

Electrocardiogram (ECG) is the most widely used diagnostic tool to monitor the condition of the cardiovascular system. Deep neural networks (DNNs), have been developed in many research labs for automatic interpretation of ECG signals to identify potential abnormalities in patient hearts. Studies have shown that given a sufficiently large amount of data, the classification accuracy of DNNs could reach human-expert cardiologist level. However, despite of the excellent performance in classification accuracy, it has been shown that DNNs are highly vulnerable to adversarial noises which are subtle changes in input of a DNN and lead to a wrong class-label prediction with a high confidence. Thus, it is challenging and essential to improve robustness of DNNs against adversarial noises for ECG signal classification -a life-critical application. In this work, we designed a CNN for classification of 12-lead ECG signals with variable length, and we applied three defense methods to improve robustness of this CNN for this classification task. The ECG data in this study is very challenging because the sample size is limited, and the length of each ECG recording varies in a large range. The evaluation results show that our customized CNN reached satisfying F1 score and average accuracy, comparable to the top-6 entries in the CPSC2018 ECG classification challenge, and the defense methods enhanced robustness of our CNN against adversarial noises and white noises, with a minimal reduction in accuracy on clean data.

Authors

Linhai Ma

1 papers

Liang Liang

1 papers

Enhance CNN Robustness Against Noises for Classification of 12-Lead ECG with Variable Length

TL;DR

Abstract

Authors

References23 items

Improve robustness of DNN for ECG signal classification: a noise-to-signal ratio perspective

On Adaptive Attacks to Adversarial Example Defenses

Adversarial Training for Free!

Theoretically Principled Trade-off between Robustness and Accuracy

MMA Training: Direct Input Space Margin Maximization through Adversarial Training

An Open Access Database for Evaluating the Algorithms of Electrocardiogram Rhythm and Morphology Abnormality Detection

ECG Heartbeat Classification: A Deep Transferable Representation

Improving DNN Robustness to Adversarial Attacks using Jacobian Regularization

Group Normalization

Adversarial Risk and the Dangers of Evaluating Against Weak Attacks

Threat of Adversarial Attacks on Deep Learning in Computer Vision: A Survey

Improving the Adversarial Robustness and Interpretability of Deep Neural Networks by Regularizing their Input Gradients

A deep convolutional neural network model to classify heartbeats

Cardiologist-Level Arrhythmia Detection with Convolutional Neural Networks

Towards Deep Learning Models Resistant to Adversarial Attacks

Adversarial examples in the physical world

Real-Time Patient-Specific ECG Classification by 1-D Convolutional Neural Networks

Practical Black-Box Attacks against Machine Learning

Classification of ECG signals using machine learning techniques: A survey

Explaining and Harnessing Adversarial Examples

Scikit-learn: Machine Learning in Python

IntelliVue Patient Monitor and Information Center

“12 -Lead ECG Monitoring TM Lead System with EASI.”

Field of Study

Journal Information

Name

Page

Venue Information

Name

Type

URL

Alternate Names