A novel tool for the removal of muscle artefacts from EEG: Improving data quality in the gamma frequency range.

Gamma activity High-frequency neuronal oscillations Muscle artefacts

Journal

Journal of neuroscience methods
ISSN: 1872-678X
Titre abrégé: J Neurosci Methods
Pays: Netherlands
ID NLM: 7905558

Informations de publication

Date de publication:
01 07 2021
Historique:
received: 25 11 2020
revised: 07 04 2021
accepted: 02 05 2021
pubmed: 9 5 2021
medline: 1 7 2021
entrez: 8 5 2021
Statut: ppublish

Résumé

The past two decades have seen a particular focus towards high-frequency neural activity in the gamma band (>30 Hz). However, gamma band activity shares frequency range with unwanted artefacts from muscular activity. We developed a novel approach to remove muscle artefacts from neurophysiological data. We re-analysed existing EEG data that were decomposed by a blind source separation method (independent component analysis, ICA), which helped to better spatially and temporally separate single muscle spikes. We then applied an adapting algorithm that detects these singled-out muscle spikes. We obtained data almost free from muscle artefacts; we needed to remove significantly fewer artefact components from the ICA and we included more trials for the statistical analysis compared to standard ICA artefact removal. All pain-related cortical effects in the gamma band have been preserved, which underlines the high efficacy and precision of this algorithm. Our results show a significant improvement of data quality by preserving task-relevant gamma oscillations of presumed cortical origin. We were able to precisely detect, gauge, and carve out single muscle spikes from the time course of neurophysiological measures without perturbing cortical gamma. We advocate the application of the tool for studies investigating gamma activity that contain a rather low number of trials, as well as for data that are highly contaminated with muscle artefacts. This validation of our tool allows for the application on event-free continuous EEG, for which the artefact removal is more challenging.

Sections du résumé

BACKGROUND
The past two decades have seen a particular focus towards high-frequency neural activity in the gamma band (>30 Hz). However, gamma band activity shares frequency range with unwanted artefacts from muscular activity.
NEW METHOD
We developed a novel approach to remove muscle artefacts from neurophysiological data. We re-analysed existing EEG data that were decomposed by a blind source separation method (independent component analysis, ICA), which helped to better spatially and temporally separate single muscle spikes. We then applied an adapting algorithm that detects these singled-out muscle spikes.
RESULTS
We obtained data almost free from muscle artefacts; we needed to remove significantly fewer artefact components from the ICA and we included more trials for the statistical analysis compared to standard ICA artefact removal. All pain-related cortical effects in the gamma band have been preserved, which underlines the high efficacy and precision of this algorithm.
CONCLUSIONS
Our results show a significant improvement of data quality by preserving task-relevant gamma oscillations of presumed cortical origin. We were able to precisely detect, gauge, and carve out single muscle spikes from the time course of neurophysiological measures without perturbing cortical gamma. We advocate the application of the tool for studies investigating gamma activity that contain a rather low number of trials, as well as for data that are highly contaminated with muscle artefacts. This validation of our tool allows for the application on event-free continuous EEG, for which the artefact removal is more challenging.

Identifiants

pubmed: 33964345
pii: S0165-0270(21)00152-7
doi: 10.1016/j.jneumeth.2021.109217
pii:
doi:

Types de publication

Journal Article

Langues

eng

Sous-ensembles de citation

IM

Pagination

109217

Informations de copyright

Copyright © 2021 Elsevier B.V. All rights reserved.

Auteurs

Alina Pauline Liebisch (AP)

Department of Neurology, Ludwig-Maximilians-Universität München, Munich, Germany.

Thomas Eggert (T)

Department of Neurology, Ludwig-Maximilians-Universität München, Munich, Germany.

Alina Shindy (A)

Department of Neurology, Ludwig-Maximilians-Universität München, Munich, Germany.

Elia Valentini (E)

Department of Psychology and Centre for Brain Science, University of Essex, Colchester, UK.

Stephanie Irving (S)

Department of Neurology, Ludwig-Maximilians-Universität München, Munich, Germany.

Anne Stankewitz (A)

Department of Neurology, Ludwig-Maximilians-Universität München, Munich, Germany.

Enrico Schulz (E)

Department of Neurology, Ludwig-Maximilians-Universität München, Munich, Germany; Department of Medical Psychology, Ludwig-Maximilians-Universität München, Munich, Germany. Electronic address: es@pain.sc.

Articles similaires

Selecting optimal software code descriptors-The case of Java.

Yegor Bugayenko, Zamira Kholmatova, Artem Kruglov et al.
1.00
Software Algorithms Programming Languages
1.00
Humans Magnetic Resonance Imaging Brain Infant, Newborn Infant, Premature
Humans Algorithms Software Artificial Intelligence Computer Simulation

Unsupervised learning for real-time and continuous gait phase detection.

Dollaporn Anopas, Yodchanan Wongsawat, Jetsada Arnin
1.00
Humans Gait Neural Networks, Computer Unsupervised Machine Learning Walking

Classifications MeSH