Channel Selection Improves MEG-based Brain-Computer Interface

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

2 Citations (Scopus)

Abstract

This study investigates the effect of channel selection on the performance of a Magnetoencephalography (MEG)-based brain-computer interface (BCI) system in terms of classification accuracy (CA). Although many efforts are currently being undertaken to develop BCI using MEG, the major concern still is low accuracy. MEG systems involve data recording from a large number of channels which may provide a better spatio-temporal resolution for assessing brain patterns, however, a large numbers of channels result in a large number of features, which further make feature learning a challenging task. In this study, we evaluated the performance of two state-of-the-art channel selection methods, i.e. class-correlation (CC) and ReliefF (RF) across six binary classification tasks with a MEG dataset of 15 healthy participants. Both CC and RF methods provided a statistically significant increase in the CA (range: 20.91 - 24.22%) compared to baseline (i.e. using 204 channels) with bandpower features from the alpha (8-12 Hz) and beta frequency bands (13-30 Hz). Moreover, both methods reduce the optimum number of channels significantly (from 204 to the range of 1-22). Reducing the number of features can significantly reduce the computational cost and increase the chances of numerical stability which are key considerations in neurofeedback (online) applications.

LanguageEnglish
Title of host publication9th International IEEE EMBS Conference on Neural Engineering, NER 2019
Place of PublicationUnited States
Pages295-298
Number of pages4
Volume2019-March
ISBN (Electronic)9781538679210
DOIs
Publication statusPublished - 20 May 2019
Event2019 9th International IEEE/EMBS Conference on Neural Engineering (NER) - San Francisco, CA, United States
Duration: 20 Mar 201823 Mar 2019
https://ieeexplore.ieee.org/xpl/conhome/8712471/proceeding

Publication series

NameInternational IEEE/EMBS Conference on Neural Engineering
ISSN (Print)1948-3546
ISSN (Electronic)1948-3554

Conference

Conference2019 9th International IEEE/EMBS Conference on Neural Engineering (NER)
CountryUnited States
CitySan Francisco, CA
Period20/03/1823/03/19
Internet address

Fingerprint

Magnetoencephalography
Brain computer interface
Data recording
Convergence of numerical methods
Frequency bands
Brain
Costs

Keywords

  • Channel Selection
  • MEG
  • BCI
  • class-correlation (CC)
  • ReliefF (RF)

Cite this

Roy, S., Rathee, D., McCreadie, K., & Prasad, G. (2019). Channel Selection Improves MEG-based Brain-Computer Interface. In 9th International IEEE EMBS Conference on Neural Engineering, NER 2019 (Vol. 2019-March, pp. 295-298). [8716948] (International IEEE/EMBS Conference on Neural Engineering). United States. https://doi.org/10.1109/NER.2019.8716948
Roy, Sujit ; Rathee, Dheeraj ; McCreadie, Karl ; Prasad, Girijesh. / Channel Selection Improves MEG-based Brain-Computer Interface. 9th International IEEE EMBS Conference on Neural Engineering, NER 2019. Vol. 2019-March United States, 2019. pp. 295-298 (International IEEE/EMBS Conference on Neural Engineering).
@inproceedings{7c1fc065ddd349d39ba044436de1b507,
title = "Channel Selection Improves MEG-based Brain-Computer Interface",
abstract = "This study investigates the effect of channel selection on the performance of a Magnetoencephalography (MEG)-based brain-computer interface (BCI) system in terms of classification accuracy (CA). Although many efforts are currently being undertaken to develop BCI using MEG, the major concern still is low accuracy. MEG systems involve data recording from a large number of channels which may provide a better spatio-temporal resolution for assessing brain patterns, however, a large numbers of channels result in a large number of features, which further make feature learning a challenging task. In this study, we evaluated the performance of two state-of-the-art channel selection methods, i.e. class-correlation (CC) and ReliefF (RF) across six binary classification tasks with a MEG dataset of 15 healthy participants. Both CC and RF methods provided a statistically significant increase in the CA (range: 20.91 - 24.22{\%}) compared to baseline (i.e. using 204 channels) with bandpower features from the alpha (8-12 Hz) and beta frequency bands (13-30 Hz). Moreover, both methods reduce the optimum number of channels significantly (from 204 to the range of 1-22). Reducing the number of features can significantly reduce the computational cost and increase the chances of numerical stability which are key considerations in neurofeedback (online) applications.",
keywords = "Channel Selection, MEG, BCI, class-correlation (CC), ReliefF (RF)",
author = "Sujit Roy and Dheeraj Rathee and Karl McCreadie and Girijesh Prasad",
year = "2019",
month = "5",
day = "20",
doi = "10.1109/NER.2019.8716948",
language = "English",
volume = "2019-March",
series = "International IEEE/EMBS Conference on Neural Engineering",
pages = "295--298",
booktitle = "9th International IEEE EMBS Conference on Neural Engineering, NER 2019",

}

Roy, S, Rathee, D, McCreadie, K & Prasad, G 2019, Channel Selection Improves MEG-based Brain-Computer Interface. in 9th International IEEE EMBS Conference on Neural Engineering, NER 2019. vol. 2019-March, 8716948, International IEEE/EMBS Conference on Neural Engineering, United States, pp. 295-298, 2019 9th International IEEE/EMBS Conference on Neural Engineering (NER), San Francisco, CA, United States, 20/03/18. https://doi.org/10.1109/NER.2019.8716948

Channel Selection Improves MEG-based Brain-Computer Interface. / Roy, Sujit; Rathee, Dheeraj; McCreadie, Karl; Prasad, Girijesh.

9th International IEEE EMBS Conference on Neural Engineering, NER 2019. Vol. 2019-March United States, 2019. p. 295-298 8716948 (International IEEE/EMBS Conference on Neural Engineering).

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

TY - GEN

T1 - Channel Selection Improves MEG-based Brain-Computer Interface

AU - Roy, Sujit

AU - Rathee, Dheeraj

AU - McCreadie, Karl

AU - Prasad, Girijesh

PY - 2019/5/20

Y1 - 2019/5/20

N2 - This study investigates the effect of channel selection on the performance of a Magnetoencephalography (MEG)-based brain-computer interface (BCI) system in terms of classification accuracy (CA). Although many efforts are currently being undertaken to develop BCI using MEG, the major concern still is low accuracy. MEG systems involve data recording from a large number of channels which may provide a better spatio-temporal resolution for assessing brain patterns, however, a large numbers of channels result in a large number of features, which further make feature learning a challenging task. In this study, we evaluated the performance of two state-of-the-art channel selection methods, i.e. class-correlation (CC) and ReliefF (RF) across six binary classification tasks with a MEG dataset of 15 healthy participants. Both CC and RF methods provided a statistically significant increase in the CA (range: 20.91 - 24.22%) compared to baseline (i.e. using 204 channels) with bandpower features from the alpha (8-12 Hz) and beta frequency bands (13-30 Hz). Moreover, both methods reduce the optimum number of channels significantly (from 204 to the range of 1-22). Reducing the number of features can significantly reduce the computational cost and increase the chances of numerical stability which are key considerations in neurofeedback (online) applications.

AB - This study investigates the effect of channel selection on the performance of a Magnetoencephalography (MEG)-based brain-computer interface (BCI) system in terms of classification accuracy (CA). Although many efforts are currently being undertaken to develop BCI using MEG, the major concern still is low accuracy. MEG systems involve data recording from a large number of channels which may provide a better spatio-temporal resolution for assessing brain patterns, however, a large numbers of channels result in a large number of features, which further make feature learning a challenging task. In this study, we evaluated the performance of two state-of-the-art channel selection methods, i.e. class-correlation (CC) and ReliefF (RF) across six binary classification tasks with a MEG dataset of 15 healthy participants. Both CC and RF methods provided a statistically significant increase in the CA (range: 20.91 - 24.22%) compared to baseline (i.e. using 204 channels) with bandpower features from the alpha (8-12 Hz) and beta frequency bands (13-30 Hz). Moreover, both methods reduce the optimum number of channels significantly (from 204 to the range of 1-22). Reducing the number of features can significantly reduce the computational cost and increase the chances of numerical stability which are key considerations in neurofeedback (online) applications.

KW - Channel Selection

KW - MEG

KW - BCI

KW - class-correlation (CC)

KW - ReliefF (RF)

UR - http://www.scopus.com/inward/record.url?scp=85066763073&partnerID=8YFLogxK

U2 - 10.1109/NER.2019.8716948

DO - 10.1109/NER.2019.8716948

M3 - Conference contribution

VL - 2019-March

T3 - International IEEE/EMBS Conference on Neural Engineering

SP - 295

EP - 298

BT - 9th International IEEE EMBS Conference on Neural Engineering, NER 2019

CY - United States

ER -

Roy S, Rathee D, McCreadie K, Prasad G. Channel Selection Improves MEG-based Brain-Computer Interface. In 9th International IEEE EMBS Conference on Neural Engineering, NER 2019. Vol. 2019-March. United States. 2019. p. 295-298. 8716948. (International IEEE/EMBS Conference on Neural Engineering). https://doi.org/10.1109/NER.2019.8716948