Deep learning models for electrocardiograms are susceptible to adversarial attack.

Nature Medicine
Xintian HanRajesh Ranganath

Abstract

Electrocardiogram (ECG) acquisition is increasingly widespread in medical and commercial devices, necessitating the development of automated interpretation strategies. Recently, deep neural networks have been used to automatically analyze ECG tracings and outperform physicians in detecting certain rhythm irregularities1. However, deep learning classifiers are susceptible to adversarial examples, which are created from raw data to fool the classifier such that it assigns the example to the wrong class, but which are undetectable to the human eye2,3. Adversarial examples have also been created for medical-related tasks4,5. However, traditional attack methods to create adversarial examples do not extend directly to ECG signals, as such methods introduce square-wave artefacts that are not physiologically plausible. Here we develop a method to construct smoothed adversarial examples for ECG tracings that are invisible to human expert evaluation and show that a deep learning model for arrhythmia detection from single-lead ECG6 is vulnerable to this type of attack. Moreover, we provide a general technique for collating and perturbing known adversarial examples to create multiple new ones. The susceptibility of deep learning ECG algori...Continue Reading

References

Apr 3, 2003·IEEE Transactions on Bio-medical Engineering·Patrick E McSharryLeonard A Smith
Nov 13, 2013·Progress in Cardiovascular Diseases·Harold L Kennedy
Mar 23, 2019·Science·Samuel G FinlaysonIsaac S Kohane

❮ Previous
Next ❯

Citations

Jul 7, 2020·Circulation. Arrhythmia and Electrophysiology·Albert K FeenyPaul J Wang
Jan 4, 2021·European Heart Journal·Charalambos AntoniadesPanos Vardas
Feb 11, 2021·Europace : European Pacing, Arrhythmias, and Cardiac Electrophysiology : Journal of the Working Groups on Cardiac Pacing, Arrhythmias, and Cardiac Cellular Electrophysiology of the European Society of Cardiology·Sulaiman SomaniBenjamin S Glicksberg
Feb 14, 2021·Scientific Reports·Cristina RuedaAdrian Lamela
Mar 14, 2021·Journal of Clinical Anesthesia·Joseph C Goldstein, Heidi V Goldstein
Aug 27, 2021·Nature Communications·Emilly M LimaAntonio Luiz P Ribeiro
Sep 18, 2021·European Heart Journal·Zachi I AttiaPaul A Friedman

❮ Previous
Next ❯

Software Mentioned

FGSM
Alivecor

Related Concepts

Related Feeds

Atrial Fibrillation

Atrial fibrillation is a common arrhythmia that is associated with substantial morbidity and mortality, particularly due to stroke and thromboembolism. Here is the latest research.

Arrhythmia

Arrhythmias are abnormalities in heart rhythms, which can be either too fast or too slow. They can result from abnormalities of the initiation of an impulse or impulse conduction or a combination of both. Here is the latest research on arrhythmias.

Related Papers

IEEE Transactions on Image Processing : a Publication of the IEEE Signal Processing Society
Yonggang ZhangDacheng Tao
IEEE Transactions on Neural Networks and Learning Systems
Xiaoyong YuanXiaolin Li
IEEE Transactions on Visualization and Computer Graphics
Kelei CaoShixia Liu
IEEE Transactions on Pattern Analysis and Machine Intelligence
Sanli TangJie Yang
© 2021 Meta ULC. All rights reserved