Deep neural networks for ECG-based pulse detection during out-of-hospital cardiac arrest

Andoni Elola, Elisabete Aramendi, Unai Irusta, Artzai Picón, Erik Alonso, Pamela Owens, Ahamed H Idris

Research output: Contribution to journalArticle

1 Citation (Scopus)

Abstract

The automatic detection of pulse during out-of-hospital cardiac arrest (OHCA) is necessary for the early recognition of the arrest and the detection of return of spontaneous circulation (end of the arrest). The only signal available in every single defibrillator and valid for the detection of pulse is the electrocardiogram (ECG). In this study we propose two deep neural network (DNN) architectures to detect pulse using short ECG segments (5 s), i.e., to classify the rhythm into pulseless electrical activity (PEA) or pulse-generating rhythm (PR). A total of 3914 5-s ECG segments, 2372 PR and 1542 PEA, were extracted from 279 OHCA episodes. Data were partitioned patient-wise into training (80%) and test (20%) sets. The first DNN architecture was a fully convolutional neural network, and the second architecture added a recurrent layer to learn temporal dependencies. Both DNN architectures were tuned using Bayesian optimization, and the results for the test set were compared to state-of-the art PR/PEA discrimination algorithms based on machine learning and hand crafted features. The PR/PEA classifiers were evaluated in terms of sensitivity (Se) for PR, specificity (Sp) for PEA, and the balanced accuracy (BAC), the average of Se and Sp. The Se/Sp/BAC of the DNN architectures were 94.1%/92.9%/93.5% for the first one, and 95.5%/91.6%/93.5% for the second one. Both architectures improved the performance of state of the art methods by more than 1.5 points in BAC.

Original languageEnglish (US)
Article number305
JournalEntropy
Volume21
Issue number3
DOIs
StatePublished - Mar 1 2019

Fingerprint

electrocardiography
rhythm
pulses
sensitivity
machine learning
classifiers
discrimination
education
optimization

Keywords

  • Bayesian optimization
  • Convolutional neural network
  • Deep learning
  • ECG
  • Out-of-hospital cardiac arrest
  • Pulse detection
  • Pulseless electrical activity

ASJC Scopus subject areas

  • Physics and Astronomy(all)

Cite this

Elola, A., Aramendi, E., Irusta, U., Picón, A., Alonso, E., Owens, P., & Idris, A. H. (2019). Deep neural networks for ECG-based pulse detection during out-of-hospital cardiac arrest. Entropy, 21(3), [305]. https://doi.org/10.3390/e21030305

Deep neural networks for ECG-based pulse detection during out-of-hospital cardiac arrest. / Elola, Andoni; Aramendi, Elisabete; Irusta, Unai; Picón, Artzai; Alonso, Erik; Owens, Pamela; Idris, Ahamed H.

In: Entropy, Vol. 21, No. 3, 305, 01.03.2019.

Research output: Contribution to journalArticle

Elola, A, Aramendi, E, Irusta, U, Picón, A, Alonso, E, Owens, P & Idris, AH 2019, 'Deep neural networks for ECG-based pulse detection during out-of-hospital cardiac arrest', Entropy, vol. 21, no. 3, 305. https://doi.org/10.3390/e21030305
Elola A, Aramendi E, Irusta U, Picón A, Alonso E, Owens P et al. Deep neural networks for ECG-based pulse detection during out-of-hospital cardiac arrest. Entropy. 2019 Mar 1;21(3). 305. https://doi.org/10.3390/e21030305
Elola, Andoni ; Aramendi, Elisabete ; Irusta, Unai ; Picón, Artzai ; Alonso, Erik ; Owens, Pamela ; Idris, Ahamed H. / Deep neural networks for ECG-based pulse detection during out-of-hospital cardiac arrest. In: Entropy. 2019 ; Vol. 21, No. 3.
@article{e6ca3b91e6e24b8ca0e0b4d8cc571cf7,
title = "Deep neural networks for ECG-based pulse detection during out-of-hospital cardiac arrest",
abstract = "The automatic detection of pulse during out-of-hospital cardiac arrest (OHCA) is necessary for the early recognition of the arrest and the detection of return of spontaneous circulation (end of the arrest). The only signal available in every single defibrillator and valid for the detection of pulse is the electrocardiogram (ECG). In this study we propose two deep neural network (DNN) architectures to detect pulse using short ECG segments (5 s), i.e., to classify the rhythm into pulseless electrical activity (PEA) or pulse-generating rhythm (PR). A total of 3914 5-s ECG segments, 2372 PR and 1542 PEA, were extracted from 279 OHCA episodes. Data were partitioned patient-wise into training (80{\%}) and test (20{\%}) sets. The first DNN architecture was a fully convolutional neural network, and the second architecture added a recurrent layer to learn temporal dependencies. Both DNN architectures were tuned using Bayesian optimization, and the results for the test set were compared to state-of-the art PR/PEA discrimination algorithms based on machine learning and hand crafted features. The PR/PEA classifiers were evaluated in terms of sensitivity (Se) for PR, specificity (Sp) for PEA, and the balanced accuracy (BAC), the average of Se and Sp. The Se/Sp/BAC of the DNN architectures were 94.1{\%}/92.9{\%}/93.5{\%} for the first one, and 95.5{\%}/91.6{\%}/93.5{\%} for the second one. Both architectures improved the performance of state of the art methods by more than 1.5 points in BAC.",
keywords = "Bayesian optimization, Convolutional neural network, Deep learning, ECG, Out-of-hospital cardiac arrest, Pulse detection, Pulseless electrical activity",
author = "Andoni Elola and Elisabete Aramendi and Unai Irusta and Artzai Pic{\'o}n and Erik Alonso and Pamela Owens and Idris, {Ahamed H}",
year = "2019",
month = "3",
day = "1",
doi = "10.3390/e21030305",
language = "English (US)",
volume = "21",
journal = "Entropy",
issn = "1099-4300",
publisher = "Multidisciplinary Digital Publishing Institute (MDPI)",
number = "3",

}

TY - JOUR

T1 - Deep neural networks for ECG-based pulse detection during out-of-hospital cardiac arrest

AU - Elola, Andoni

AU - Aramendi, Elisabete

AU - Irusta, Unai

AU - Picón, Artzai

AU - Alonso, Erik

AU - Owens, Pamela

AU - Idris, Ahamed H

PY - 2019/3/1

Y1 - 2019/3/1

N2 - The automatic detection of pulse during out-of-hospital cardiac arrest (OHCA) is necessary for the early recognition of the arrest and the detection of return of spontaneous circulation (end of the arrest). The only signal available in every single defibrillator and valid for the detection of pulse is the electrocardiogram (ECG). In this study we propose two deep neural network (DNN) architectures to detect pulse using short ECG segments (5 s), i.e., to classify the rhythm into pulseless electrical activity (PEA) or pulse-generating rhythm (PR). A total of 3914 5-s ECG segments, 2372 PR and 1542 PEA, were extracted from 279 OHCA episodes. Data were partitioned patient-wise into training (80%) and test (20%) sets. The first DNN architecture was a fully convolutional neural network, and the second architecture added a recurrent layer to learn temporal dependencies. Both DNN architectures were tuned using Bayesian optimization, and the results for the test set were compared to state-of-the art PR/PEA discrimination algorithms based on machine learning and hand crafted features. The PR/PEA classifiers were evaluated in terms of sensitivity (Se) for PR, specificity (Sp) for PEA, and the balanced accuracy (BAC), the average of Se and Sp. The Se/Sp/BAC of the DNN architectures were 94.1%/92.9%/93.5% for the first one, and 95.5%/91.6%/93.5% for the second one. Both architectures improved the performance of state of the art methods by more than 1.5 points in BAC.

AB - The automatic detection of pulse during out-of-hospital cardiac arrest (OHCA) is necessary for the early recognition of the arrest and the detection of return of spontaneous circulation (end of the arrest). The only signal available in every single defibrillator and valid for the detection of pulse is the electrocardiogram (ECG). In this study we propose two deep neural network (DNN) architectures to detect pulse using short ECG segments (5 s), i.e., to classify the rhythm into pulseless electrical activity (PEA) or pulse-generating rhythm (PR). A total of 3914 5-s ECG segments, 2372 PR and 1542 PEA, were extracted from 279 OHCA episodes. Data were partitioned patient-wise into training (80%) and test (20%) sets. The first DNN architecture was a fully convolutional neural network, and the second architecture added a recurrent layer to learn temporal dependencies. Both DNN architectures were tuned using Bayesian optimization, and the results for the test set were compared to state-of-the art PR/PEA discrimination algorithms based on machine learning and hand crafted features. The PR/PEA classifiers were evaluated in terms of sensitivity (Se) for PR, specificity (Sp) for PEA, and the balanced accuracy (BAC), the average of Se and Sp. The Se/Sp/BAC of the DNN architectures were 94.1%/92.9%/93.5% for the first one, and 95.5%/91.6%/93.5% for the second one. Both architectures improved the performance of state of the art methods by more than 1.5 points in BAC.

KW - Bayesian optimization

KW - Convolutional neural network

KW - Deep learning

KW - ECG

KW - Out-of-hospital cardiac arrest

KW - Pulse detection

KW - Pulseless electrical activity

UR - http://www.scopus.com/inward/record.url?scp=85063594590&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85063594590&partnerID=8YFLogxK

U2 - 10.3390/e21030305

DO - 10.3390/e21030305

M3 - Article

AN - SCOPUS:85063594590

VL - 21

JO - Entropy

JF - Entropy

SN - 1099-4300

IS - 3

M1 - 305

ER -