Paper Info
Title
Adversarial de-noising of electrocardiogram.
Abstract
The electrocardiogram (ECG) is an important index to monitor heart health and to treat heart diseases. The ECG signals acquisition process is often accompanied by a large amount of noise, which will seriously affect the doctor’s diagnosis of patients, especially in the telemedicine environment. However, existing de-noising methods suffer from several deficiencies: (1) the local and global correlations of ECG signals are not considered comprehensively; (2) adaptability is not good enough for various noise; (3) severe distortion in signals may be triggered. In this paper, we propose an adversarial method for ECG signals de-noising. The method adopts a newly designed loss function to consider both global and local characteristics of signals, utilizes the adversarial characteristics to accumulate knowledge on the distribution of ECG noise continuously through the game between the generator and the discriminator, and evaluates the quality of de-noised signals against SVM algorithm. The extensive experiments show that compared to the state-of-the-art methods, our method achieves up to about 62% improvement on the SNR of de-noised signals on average. Evaluations on the quality of de-noised signals imply that our method can effectively preserve useful medical characteristics of ECG signals.
Year
DOI
Venue
2019
10.1016/j.neucom.2019.03.083
Neurocomputing
Keywords
Field
DocType
Electrocardiogram signal,Generative adversarial networks,Noise reduction
Adaptability,Discriminator,Pattern recognition,Support vector machine,Artificial intelligence,Distortion,Mathematics,Adversarial system
Journal
Volume
ISSN
Citations 
349
0925-2312
3
PageRank 
References 
Authors
0.39
0
7
Name
Order
Citations
PageRank
Jilong Wang15719.88
Renfa Li264797.10
Rui Li34911.99
Keqin Li42778242.13
Haibo Zeng5328.74
Guoqi Xie610515.41
Li Liu716950.09