Automatic 1D convolutional neural network-based detection of artifacts in MEG acquired without electrooculography or electrocardiography

Prabhat Garg, Elizabeth Davenport, Gowtham Murugesan, Ben Wagner, Christopher Whitlow, Joseph A Maldjian, Albert Montillo

Research output: Chapter in Book/Report/Conference proceedingConference contribution

2 Citations (Scopus)

Abstract

Magnetoencephalography (MEG) is a functional neuroimaging tool that records the magnetic fields induced by electrical neuronal activity; however, signal from non-neuronal sources can corrupt the data. Eye-Blinks (EB) and Cardiac Activity (CA) are two of the most common types of non-neuronal artifacts. They can be measured by affixing eye proximal electrodes, as in electrooculography (EOG) and chest electrodes, as in electrocardiography (EKG), however this complicates imaging setup, decreases patient comfort, and often induces further artifacts from facial twitching and postural muscle movement. We propose an EOG- and EKG-free approach to identify eye-blink, cardiac, or neuronal signals for automated artifact suppression. Our contributions are two-fold. First, we combine a data driven, multivariate decomposition approach based on Independent Component Analysis (ICA) and a highly accurate classifier constructed as a deep 1-D Convolutional Neural Network. Second, we visualize the features learned to reveal what features the model uses and to bolster user confidence in our model's training and potential for generalization. We train and test three variants of our method on resting state MEG data from 49 subjects. Our cardiac model achieves a 96% sensitivity and 99% specificity on the set-aside test-set. Our eye-blink model achieves a sensitivity of 85% and specificity of 97%. This work facilitates automated MEG processing for both, clinical and research use, and can obviate the need for EOG or EKG electrodes.

Original languageEnglish (US)
Title of host publication2017 International Workshop on Pattern Recognition in Neuroimaging, PRNI 2017
PublisherInstitute of Electrical and Electronics Engineers Inc.
ISBN (Electronic)9781538631591
DOIs
StatePublished - Jul 14 2017
Event2017 International Workshop on Pattern Recognition in Neuroimaging, PRNI 2017 - Toronto, Canada
Duration: Jun 21 2017Jun 23 2017

Other

Other2017 International Workshop on Pattern Recognition in Neuroimaging, PRNI 2017
CountryCanada
CityToronto
Period6/21/176/23/17

Fingerprint

Electrooculography
Magnetoencephalography
Electrocardiography
Artifacts
Neural networks
Electrodes
Functional neuroimaging
Sensitivity and Specificity
Functional Neuroimaging
Independent component analysis
Magnetic Fields
Muscle
Classifiers
Thorax
Magnetic fields
Decomposition
Imaging techniques
Muscles
Processing
Research

Keywords

  • artifact
  • CNN
  • deep learning
  • EKG
  • EOG
  • MEG

ASJC Scopus subject areas

  • Computer Vision and Pattern Recognition
  • Radiology Nuclear Medicine and imaging
  • Neuroscience (miscellaneous)

Cite this

Garg, P., Davenport, E., Murugesan, G., Wagner, B., Whitlow, C., Maldjian, J. A., & Montillo, A. (2017). Automatic 1D convolutional neural network-based detection of artifacts in MEG acquired without electrooculography or electrocardiography. In 2017 International Workshop on Pattern Recognition in Neuroimaging, PRNI 2017 [7981506] Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/PRNI.2017.7981506

Automatic 1D convolutional neural network-based detection of artifacts in MEG acquired without electrooculography or electrocardiography. / Garg, Prabhat; Davenport, Elizabeth; Murugesan, Gowtham; Wagner, Ben; Whitlow, Christopher; Maldjian, Joseph A; Montillo, Albert.

2017 International Workshop on Pattern Recognition in Neuroimaging, PRNI 2017. Institute of Electrical and Electronics Engineers Inc., 2017. 7981506.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Garg, P, Davenport, E, Murugesan, G, Wagner, B, Whitlow, C, Maldjian, JA & Montillo, A 2017, Automatic 1D convolutional neural network-based detection of artifacts in MEG acquired without electrooculography or electrocardiography. in 2017 International Workshop on Pattern Recognition in Neuroimaging, PRNI 2017., 7981506, Institute of Electrical and Electronics Engineers Inc., 2017 International Workshop on Pattern Recognition in Neuroimaging, PRNI 2017, Toronto, Canada, 6/21/17. https://doi.org/10.1109/PRNI.2017.7981506
Garg P, Davenport E, Murugesan G, Wagner B, Whitlow C, Maldjian JA et al. Automatic 1D convolutional neural network-based detection of artifacts in MEG acquired without electrooculography or electrocardiography. In 2017 International Workshop on Pattern Recognition in Neuroimaging, PRNI 2017. Institute of Electrical and Electronics Engineers Inc. 2017. 7981506 https://doi.org/10.1109/PRNI.2017.7981506
Garg, Prabhat ; Davenport, Elizabeth ; Murugesan, Gowtham ; Wagner, Ben ; Whitlow, Christopher ; Maldjian, Joseph A ; Montillo, Albert. / Automatic 1D convolutional neural network-based detection of artifacts in MEG acquired without electrooculography or electrocardiography. 2017 International Workshop on Pattern Recognition in Neuroimaging, PRNI 2017. Institute of Electrical and Electronics Engineers Inc., 2017.
@inproceedings{f7692b4d48c142b7884fefce6967999a,
title = "Automatic 1D convolutional neural network-based detection of artifacts in MEG acquired without electrooculography or electrocardiography",
abstract = "Magnetoencephalography (MEG) is a functional neuroimaging tool that records the magnetic fields induced by electrical neuronal activity; however, signal from non-neuronal sources can corrupt the data. Eye-Blinks (EB) and Cardiac Activity (CA) are two of the most common types of non-neuronal artifacts. They can be measured by affixing eye proximal electrodes, as in electrooculography (EOG) and chest electrodes, as in electrocardiography (EKG), however this complicates imaging setup, decreases patient comfort, and often induces further artifacts from facial twitching and postural muscle movement. We propose an EOG- and EKG-free approach to identify eye-blink, cardiac, or neuronal signals for automated artifact suppression. Our contributions are two-fold. First, we combine a data driven, multivariate decomposition approach based on Independent Component Analysis (ICA) and a highly accurate classifier constructed as a deep 1-D Convolutional Neural Network. Second, we visualize the features learned to reveal what features the model uses and to bolster user confidence in our model's training and potential for generalization. We train and test three variants of our method on resting state MEG data from 49 subjects. Our cardiac model achieves a 96{\%} sensitivity and 99{\%} specificity on the set-aside test-set. Our eye-blink model achieves a sensitivity of 85{\%} and specificity of 97{\%}. This work facilitates automated MEG processing for both, clinical and research use, and can obviate the need for EOG or EKG electrodes.",
keywords = "artifact, CNN, deep learning, EKG, EOG, MEG",
author = "Prabhat Garg and Elizabeth Davenport and Gowtham Murugesan and Ben Wagner and Christopher Whitlow and Maldjian, {Joseph A} and Albert Montillo",
year = "2017",
month = "7",
day = "14",
doi = "10.1109/PRNI.2017.7981506",
language = "English (US)",
booktitle = "2017 International Workshop on Pattern Recognition in Neuroimaging, PRNI 2017",
publisher = "Institute of Electrical and Electronics Engineers Inc.",

}

TY - GEN

T1 - Automatic 1D convolutional neural network-based detection of artifacts in MEG acquired without electrooculography or electrocardiography

AU - Garg, Prabhat

AU - Davenport, Elizabeth

AU - Murugesan, Gowtham

AU - Wagner, Ben

AU - Whitlow, Christopher

AU - Maldjian, Joseph A

AU - Montillo, Albert

PY - 2017/7/14

Y1 - 2017/7/14

N2 - Magnetoencephalography (MEG) is a functional neuroimaging tool that records the magnetic fields induced by electrical neuronal activity; however, signal from non-neuronal sources can corrupt the data. Eye-Blinks (EB) and Cardiac Activity (CA) are two of the most common types of non-neuronal artifacts. They can be measured by affixing eye proximal electrodes, as in electrooculography (EOG) and chest electrodes, as in electrocardiography (EKG), however this complicates imaging setup, decreases patient comfort, and often induces further artifacts from facial twitching and postural muscle movement. We propose an EOG- and EKG-free approach to identify eye-blink, cardiac, or neuronal signals for automated artifact suppression. Our contributions are two-fold. First, we combine a data driven, multivariate decomposition approach based on Independent Component Analysis (ICA) and a highly accurate classifier constructed as a deep 1-D Convolutional Neural Network. Second, we visualize the features learned to reveal what features the model uses and to bolster user confidence in our model's training and potential for generalization. We train and test three variants of our method on resting state MEG data from 49 subjects. Our cardiac model achieves a 96% sensitivity and 99% specificity on the set-aside test-set. Our eye-blink model achieves a sensitivity of 85% and specificity of 97%. This work facilitates automated MEG processing for both, clinical and research use, and can obviate the need for EOG or EKG electrodes.

AB - Magnetoencephalography (MEG) is a functional neuroimaging tool that records the magnetic fields induced by electrical neuronal activity; however, signal from non-neuronal sources can corrupt the data. Eye-Blinks (EB) and Cardiac Activity (CA) are two of the most common types of non-neuronal artifacts. They can be measured by affixing eye proximal electrodes, as in electrooculography (EOG) and chest electrodes, as in electrocardiography (EKG), however this complicates imaging setup, decreases patient comfort, and often induces further artifacts from facial twitching and postural muscle movement. We propose an EOG- and EKG-free approach to identify eye-blink, cardiac, or neuronal signals for automated artifact suppression. Our contributions are two-fold. First, we combine a data driven, multivariate decomposition approach based on Independent Component Analysis (ICA) and a highly accurate classifier constructed as a deep 1-D Convolutional Neural Network. Second, we visualize the features learned to reveal what features the model uses and to bolster user confidence in our model's training and potential for generalization. We train and test three variants of our method on resting state MEG data from 49 subjects. Our cardiac model achieves a 96% sensitivity and 99% specificity on the set-aside test-set. Our eye-blink model achieves a sensitivity of 85% and specificity of 97%. This work facilitates automated MEG processing for both, clinical and research use, and can obviate the need for EOG or EKG electrodes.

KW - artifact

KW - CNN

KW - deep learning

KW - EKG

KW - EOG

KW - MEG

UR - http://www.scopus.com/inward/record.url?scp=85027981005&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85027981005&partnerID=8YFLogxK

U2 - 10.1109/PRNI.2017.7981506

DO - 10.1109/PRNI.2017.7981506

M3 - Conference contribution

C2 - 31656826

AN - SCOPUS:85027981005

BT - 2017 International Workshop on Pattern Recognition in Neuroimaging, PRNI 2017

PB - Institute of Electrical and Electronics Engineers Inc.

ER -