Cleaning ECG with Deep Learning: A Denoiser Based on Gated Recurrent Units

Mariana Dias; Phillip Probst; Luís Silva; Hugo Gamboa

doi:10.1007/978-3-031-36007-7_11

Cleaning ECG with Deep Learning: A Denoiser Based on Gated Recurrent Units

Mariana Dias, Phillip Probst, Luís Silva, Hugo Gamboa

LIBPhys-UNL

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

Abstract

The electrocardiogram (ECG) is an established exam to diagnose cardiovascular disease. Due to the increasing popularity of wearables, a wide part of the population has now access to (self-)monitorization of cardiovascular activity. Wearable ECG acquisition systems are prone to noise sources stemming from surrounding muscle activation, electrode movement, and baseline wander. Hence, many attempts have been made to develop algorithms that clean the signal, but their performance falls short when applied to very noisy signals. Acknowledging the demonstrated power of Deep Learning on timeseries processing, we propose a ECG denoiser based on Gated Recurrent Units (GRU). Noisy ECG samples were created by adding noise from the MIT-BIH Noise Stress Test database to ECG samples from the PTB-XL database. The trained network proves to remove various common noise types resulting in high quality ECG signals, while having a much smaller number of parameters compared to state-of-the-art DL approaches.

Original language	English
Title of host publication	Technological Innovation for Connected Cyber Physical Spaces
Subtitle of host publication	14th IFIP WG 5.5/SOCOLNET Doctoral Conference on Computing, Electrical and Industrial Systems, DoCEIS 2023, Caparica, Portugal, July 5–7, 2023, Proceedings
Editors	Luís M. Camarinha-Matos, Filipa Ferrada
Place of Publication	Cham
Publisher	Springer
Pages	149-160
Number of pages	12
ISBN (Electronic)	978-3-031-36007-7
ISBN (Print)	978-3-031-36006-0
DOIs	https://doi.org/10.1007/978-3-031-36007-7_11
Publication status	Published - 2023
Event	14th IFIP WG 5.5/SOCOLNET Advanced Doctoral Conference on Computing, Electrical and Industrial Systems, DoCEIS 2023 - Caparica, Portugal Duration: 5 Jul 2023 → 7 Jul 2023

Publication series

Name	IFIP Advances in Information and Communication Technology
Publisher	Springer
Volume	678
ISSN (Print)	1868-4238
ISSN (Electronic)	1868-422X

Conference

Conference	14th IFIP WG 5.5/SOCOLNET Advanced Doctoral Conference on Computing, Electrical and Industrial Systems, DoCEIS 2023
Country/Territory	Portugal
City	Caparica
Period	5/07/23 → 7/07/23

Keywords

Deep Learning
Denoiser
ECG
GRU
Signal Processing

UN SDGs

This output contributes to the following UN Sustainable Development Goals (SDGs)

Access to Document

10.1007/978-3-031-36007-7_11

Cite this

Dias, M., Probst, P., Silva, L., & Gamboa, H. (2023). Cleaning ECG with Deep Learning: A Denoiser Based on Gated Recurrent Units. In L. M. Camarinha-Matos, & F. Ferrada (Eds.), Technological Innovation for Connected Cyber Physical Spaces: 14th IFIP WG 5.5/SOCOLNET Doctoral Conference on Computing, Electrical and Industrial Systems, DoCEIS 2023, Caparica, Portugal, July 5–7, 2023, Proceedings (pp. 149-160). (IFIP Advances in Information and Communication Technology; Vol. 678). Springer. https://doi.org/10.1007/978-3-031-36007-7_11

Dias, Mariana ; Probst, Phillip ; Silva, Luís et al. / Cleaning ECG with Deep Learning : A Denoiser Based on Gated Recurrent Units. Technological Innovation for Connected Cyber Physical Spaces: 14th IFIP WG 5.5/SOCOLNET Doctoral Conference on Computing, Electrical and Industrial Systems, DoCEIS 2023, Caparica, Portugal, July 5–7, 2023, Proceedings. editor / Luís M. Camarinha-Matos ; Filipa Ferrada. Cham : Springer, 2023. pp. 149-160 (IFIP Advances in Information and Communication Technology).

@inproceedings{0aba5a1f210842b78a2d2f49cbdb0c46,

title = "Cleaning ECG with Deep Learning: A Denoiser Based on Gated Recurrent Units",

abstract = "The electrocardiogram (ECG) is an established exam to diagnose cardiovascular disease. Due to the increasing popularity of wearables, a wide part of the population has now access to (self-)monitorization of cardiovascular activity. Wearable ECG acquisition systems are prone to noise sources stemming from surrounding muscle activation, electrode movement, and baseline wander. Hence, many attempts have been made to develop algorithms that clean the signal, but their performance falls short when applied to very noisy signals. Acknowledging the demonstrated power of Deep Learning on timeseries processing, we propose a ECG denoiser based on Gated Recurrent Units (GRU). Noisy ECG samples were created by adding noise from the MIT-BIH Noise Stress Test database to ECG samples from the PTB-XL database. The trained network proves to remove various common noise types resulting in high quality ECG signals, while having a much smaller number of parameters compared to state-of-the-art DL approaches.",

keywords = "Deep Learning, Denoiser, ECG, GRU, Signal Processing",

author = "Mariana Dias and Phillip Probst and Lu{\'i}s Silva and Hugo Gamboa",

note = "Funding Information: info:eu-repo/grantAgreement/FCT/OE/SFRH%2FBD%2F151375%2F2021/PT# info:eu-repo/grantAgreement/FCT//PRT%2FBD%2F152843%2F2021/PT# Acknowledgments. This work was supported by Project OPERATOR (NORTE01-0247-FEDER-045910), cofinanced by the European Regional Development Fund through the North Portugal Regional Operational Program and Lisbon Regional Operational Program and by the Portuguese Foundation for Science and Technology, under the MIT Portugal Program. Publisher Copyright: {\textcopyright} 2023, The Author(s), under exclusive license to Springer Nature Switzerland AG.; 14th IFIP WG 5.5/SOCOLNET Advanced Doctoral Conference on Computing, Electrical and Industrial Systems, DoCEIS 2023 ; Conference date: 05-07-2023 Through 07-07-2023",

year = "2023",

doi = "10.1007/978-3-031-36007-7_11",

language = "English",

isbn = "978-3-031-36006-0",

series = "IFIP Advances in Information and Communication Technology",

publisher = "Springer",

pages = "149--160",

editor = "Camarinha-Matos, {Lu{\'i}s M.} and Filipa Ferrada",

booktitle = "Technological Innovation for Connected Cyber Physical Spaces",

address = "Netherlands",

}

Dias, M , Probst, P , Silva, L & Gamboa, H 2023, Cleaning ECG with Deep Learning: A Denoiser Based on Gated Recurrent Units. in LM Camarinha-Matos & F Ferrada (eds), Technological Innovation for Connected Cyber Physical Spaces: 14th IFIP WG 5.5/SOCOLNET Doctoral Conference on Computing, Electrical and Industrial Systems, DoCEIS 2023, Caparica, Portugal, July 5–7, 2023, Proceedings. IFIP Advances in Information and Communication Technology, vol. 678, Springer, Cham, pp. 149-160, 14th IFIP WG 5.5/SOCOLNET Advanced Doctoral Conference on Computing, Electrical and Industrial Systems, DoCEIS 2023, Caparica, Portugal, 5/07/23. https://doi.org/10.1007/978-3-031-36007-7_11

Cleaning ECG with Deep Learning: A Denoiser Based on Gated Recurrent Units. / Dias, Mariana ; Probst, Phillip ; Silva, Luís et al.
Technological Innovation for Connected Cyber Physical Spaces: 14th IFIP WG 5.5/SOCOLNET Doctoral Conference on Computing, Electrical and Industrial Systems, DoCEIS 2023, Caparica, Portugal, July 5–7, 2023, Proceedings. ed. / Luís M. Camarinha-Matos; Filipa Ferrada. Cham: Springer, 2023. p. 149-160 (IFIP Advances in Information and Communication Technology; Vol. 678).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - Cleaning ECG with Deep Learning

T2 - 14th IFIP WG 5.5/SOCOLNET Advanced Doctoral Conference on Computing, Electrical and Industrial Systems, DoCEIS 2023

AU - Dias, Mariana

AU - Probst, Phillip

AU - Silva, Luís

AU - Gamboa, Hugo

N1 - Funding Information: info:eu-repo/grantAgreement/FCT/OE/SFRH%2FBD%2F151375%2F2021/PT# info:eu-repo/grantAgreement/FCT//PRT%2FBD%2F152843%2F2021/PT# Acknowledgments. This work was supported by Project OPERATOR (NORTE01-0247-FEDER-045910), cofinanced by the European Regional Development Fund through the North Portugal Regional Operational Program and Lisbon Regional Operational Program and by the Portuguese Foundation for Science and Technology, under the MIT Portugal Program. Publisher Copyright: © 2023, The Author(s), under exclusive license to Springer Nature Switzerland AG.

PY - 2023

Y1 - 2023

N2 - The electrocardiogram (ECG) is an established exam to diagnose cardiovascular disease. Due to the increasing popularity of wearables, a wide part of the population has now access to (self-)monitorization of cardiovascular activity. Wearable ECG acquisition systems are prone to noise sources stemming from surrounding muscle activation, electrode movement, and baseline wander. Hence, many attempts have been made to develop algorithms that clean the signal, but their performance falls short when applied to very noisy signals. Acknowledging the demonstrated power of Deep Learning on timeseries processing, we propose a ECG denoiser based on Gated Recurrent Units (GRU). Noisy ECG samples were created by adding noise from the MIT-BIH Noise Stress Test database to ECG samples from the PTB-XL database. The trained network proves to remove various common noise types resulting in high quality ECG signals, while having a much smaller number of parameters compared to state-of-the-art DL approaches.

AB - The electrocardiogram (ECG) is an established exam to diagnose cardiovascular disease. Due to the increasing popularity of wearables, a wide part of the population has now access to (self-)monitorization of cardiovascular activity. Wearable ECG acquisition systems are prone to noise sources stemming from surrounding muscle activation, electrode movement, and baseline wander. Hence, many attempts have been made to develop algorithms that clean the signal, but their performance falls short when applied to very noisy signals. Acknowledging the demonstrated power of Deep Learning on timeseries processing, we propose a ECG denoiser based on Gated Recurrent Units (GRU). Noisy ECG samples were created by adding noise from the MIT-BIH Noise Stress Test database to ECG samples from the PTB-XL database. The trained network proves to remove various common noise types resulting in high quality ECG signals, while having a much smaller number of parameters compared to state-of-the-art DL approaches.

KW - Deep Learning

KW - Denoiser

KW - ECG

KW - GRU

KW - Signal Processing

UR - http://www.scopus.com/inward/record.url?scp=85164970284&partnerID=8YFLogxK

U2 - 10.1007/978-3-031-36007-7_11

DO - 10.1007/978-3-031-36007-7_11

M3 - Conference contribution

AN - SCOPUS:85164970284

SN - 978-3-031-36006-0

T3 - IFIP Advances in Information and Communication Technology

SP - 149

EP - 160

BT - Technological Innovation for Connected Cyber Physical Spaces

A2 - Camarinha-Matos, Luís M.

A2 - Ferrada, Filipa

PB - Springer

CY - Cham

Y2 - 5 July 2023 through 7 July 2023

ER -

Dias M , Probst P , Silva L , Gamboa H. Cleaning ECG with Deep Learning: A Denoiser Based on Gated Recurrent Units. In Camarinha-Matos LM, Ferrada F, editors, Technological Innovation for Connected Cyber Physical Spaces: 14th IFIP WG 5.5/SOCOLNET Doctoral Conference on Computing, Electrical and Industrial Systems, DoCEIS 2023, Caparica, Portugal, July 5–7, 2023, Proceedings. Cham: Springer. 2023. p. 149-160. (IFIP Advances in Information and Communication Technology). doi: 10.1007/978-3-031-36007-7_11

Cleaning ECG with Deep Learning: A Denoiser Based on Gated Recurrent Units

Abstract

Publication series

Conference

Keywords

UN SDGs

Access to Document

Other files and links

Fingerprint

Cite this