Determining the optimal number of MEG trials: A machine learning and speech decoding perspective

Debadatta Dash; Paul Ferrari; Saleem Malik; Albert Montillo; Joseph A Maldjian; Jun Wang

doi:10.1007/978-3-030-05587-5_16

Determining the optimal number of MEG trials: A machine learning and speech decoding perspective

Debadatta Dash, Paul Ferrari, Saleem Malik, Albert Montillo, Joseph A Maldjian, Jun Wang

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

12 Scopus citations

Abstract

Advancing the knowledge about neural speech mechanisms is critical for developing next-generation, faster brain computer interface to assist in speech communication for the patients with severe neurological conditions (e.g., locked-in syndrome). Among current neuroimaging techniques, Magnetoencephalography (MEG) provides direct representation for the large-scale neural dynamics of underlying cognitive processes based on its optimal spatiotemporal resolution. However, the MEG measured neural signals are smaller in magnitude compared to the background noise and hence, MEG usually suffers from a low signal-to-noise ratio (SNR) at the single-trial level. To overcome this limitation, it is common to record many trials of the same event-task and use the time-locked average signal for analysis, which can be very time consuming. In this study, we investigated the effect of the number of MEG recording trials required for speech decoding using a machine learning algorithm. We used a wavelet filter for generating the denoised neural features to train an Artificial Neural Network (ANN) for speech decoding. We found that wavelet based denoising increased the SNR of the neural signal prior to analysis and facilitated accurate speech decoding performance using as few as 40 single-trials. This study may open up the possibility of limiting MEG trials for other task evoked studies as well.

Original language	English (US)
Title of host publication	Brain Informatics - International Conference, BI 2018, Proceedings
Editors	Yang Yang, Vicky Yamamoto, Shouyi Wang, Erick Jones, Jianzhong Su, Tom Mitchell, Leon Iasemidis
Publisher	Springer Verlag
Pages	163-172
Number of pages	10
ISBN (Print)	9783030055868
DOIs	https://doi.org/10.1007/978-3-030-05587-5_16
State	Published - 2018
Event	International Conference on Brain Informatics, BI 2018 - Arlington, United States Duration: Dec 7 2018 → Dec 9 2018

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	11309 LNAI
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Other

Other	International Conference on Brain Informatics, BI 2018
Country/Territory	United States
City	Arlington
Period	12/7/18 → 12/9/18

Keywords

Artificial neural network
MEG
Speech
Wavelets

ASJC Scopus subject areas

Theoretical Computer Science
General Computer Science

Access to Document

10.1007/978-3-030-05587-5_16

Cite this

Dash, D., Ferrari, P., Malik, S., Montillo, A., Maldjian, J. A., & Wang, J. (2018). Determining the optimal number of MEG trials: A machine learning and speech decoding perspective. In Y. Yang, V. Yamamoto, S. Wang, E. Jones, J. Su, T. Mitchell, & L. Iasemidis (Eds.), Brain Informatics - International Conference, BI 2018, Proceedings (pp. 163-172). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 11309 LNAI). Springer Verlag. https://doi.org/10.1007/978-3-030-05587-5_16

Determining the optimal number of MEG trials: A machine learning and speech decoding perspective. / Dash, Debadatta; Ferrari, Paul; Malik, Saleem et al.
Brain Informatics - International Conference, BI 2018, Proceedings. ed. / Yang Yang; Vicky Yamamoto; Shouyi Wang; Erick Jones; Jianzhong Su; Tom Mitchell; Leon Iasemidis. Springer Verlag, 2018. p. 163-172 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 11309 LNAI).

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

Dash, D, Ferrari, P, Malik, S, Montillo, A , Maldjian, JA & Wang, J 2018, Determining the optimal number of MEG trials: A machine learning and speech decoding perspective. in Y Yang, V Yamamoto, S Wang, E Jones, J Su, T Mitchell & L Iasemidis (eds), Brain Informatics - International Conference, BI 2018, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 11309 LNAI, Springer Verlag, pp. 163-172, International Conference on Brain Informatics, BI 2018, Arlington, United States, 12/7/18. https://doi.org/10.1007/978-3-030-05587-5_16

Dash D, Ferrari P, Malik S, Montillo A , Maldjian JA, Wang J. Determining the optimal number of MEG trials: A machine learning and speech decoding perspective. In Yang Y, Yamamoto V, Wang S, Jones E, Su J, Mitchell T, Iasemidis L, editors, Brain Informatics - International Conference, BI 2018, Proceedings. Springer Verlag. 2018. p. 163-172. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/978-3-030-05587-5_16

Dash, Debadatta ; Ferrari, Paul ; Malik, Saleem et al. / Determining the optimal number of MEG trials : A machine learning and speech decoding perspective. Brain Informatics - International Conference, BI 2018, Proceedings. editor / Yang Yang ; Vicky Yamamoto ; Shouyi Wang ; Erick Jones ; Jianzhong Su ; Tom Mitchell ; Leon Iasemidis. Springer Verlag, 2018. pp. 163-172 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{0fff4a2ae8994bc5a6ec919d1a47d9b5,

title = "Determining the optimal number of MEG trials: A machine learning and speech decoding perspective",

abstract = "Advancing the knowledge about neural speech mechanisms is critical for developing next-generation, faster brain computer interface to assist in speech communication for the patients with severe neurological conditions (e.g., locked-in syndrome). Among current neuroimaging techniques, Magnetoencephalography (MEG) provides direct representation for the large-scale neural dynamics of underlying cognitive processes based on its optimal spatiotemporal resolution. However, the MEG measured neural signals are smaller in magnitude compared to the background noise and hence, MEG usually suffers from a low signal-to-noise ratio (SNR) at the single-trial level. To overcome this limitation, it is common to record many trials of the same event-task and use the time-locked average signal for analysis, which can be very time consuming. In this study, we investigated the effect of the number of MEG recording trials required for speech decoding using a machine learning algorithm. We used a wavelet filter for generating the denoised neural features to train an Artificial Neural Network (ANN) for speech decoding. We found that wavelet based denoising increased the SNR of the neural signal prior to analysis and facilitated accurate speech decoding performance using as few as 40 single-trials. This study may open up the possibility of limiting MEG trials for other task evoked studies as well.",

keywords = "Artificial neural network, MEG, Speech, Wavelets",

author = "Debadatta Dash and Paul Ferrari and Saleem Malik and Albert Montillo and Maldjian, {Joseph A} and Jun Wang",

note = "Funding Information: Acknowledgment. This project was supported by the University of Texas System Brain Research grant and by the National Institutes of Health (NIH) under award number R03 DC013990. We thank Dr. Mark McManis, Dr. Ted Mau, Dr. Angel W. Hernandez-Mulero, Kanishk Goel and the volunteering participants. Publisher Copyright: {\textcopyright} 2018, Springer Nature Switzerland AG.; International Conference on Brain Informatics, BI 2018 ; Conference date: 07-12-2018 Through 09-12-2018",

year = "2018",

doi = "10.1007/978-3-030-05587-5_16",

language = "English (US)",

isbn = "9783030055868",

series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

publisher = "Springer Verlag",

pages = "163--172",

editor = "Yang Yang and Vicky Yamamoto and Shouyi Wang and Erick Jones and Jianzhong Su and Tom Mitchell and Leon Iasemidis",

booktitle = "Brain Informatics - International Conference, BI 2018, Proceedings",

}

TY - GEN

T1 - Determining the optimal number of MEG trials

T2 - International Conference on Brain Informatics, BI 2018

AU - Dash, Debadatta

AU - Ferrari, Paul

AU - Malik, Saleem

AU - Montillo, Albert

AU - Maldjian, Joseph A

AU - Wang, Jun

N1 - Funding Information: Acknowledgment. This project was supported by the University of Texas System Brain Research grant and by the National Institutes of Health (NIH) under award number R03 DC013990. We thank Dr. Mark McManis, Dr. Ted Mau, Dr. Angel W. Hernandez-Mulero, Kanishk Goel and the volunteering participants. Publisher Copyright: © 2018, Springer Nature Switzerland AG.

PY - 2018

Y1 - 2018

N2 - Advancing the knowledge about neural speech mechanisms is critical for developing next-generation, faster brain computer interface to assist in speech communication for the patients with severe neurological conditions (e.g., locked-in syndrome). Among current neuroimaging techniques, Magnetoencephalography (MEG) provides direct representation for the large-scale neural dynamics of underlying cognitive processes based on its optimal spatiotemporal resolution. However, the MEG measured neural signals are smaller in magnitude compared to the background noise and hence, MEG usually suffers from a low signal-to-noise ratio (SNR) at the single-trial level. To overcome this limitation, it is common to record many trials of the same event-task and use the time-locked average signal for analysis, which can be very time consuming. In this study, we investigated the effect of the number of MEG recording trials required for speech decoding using a machine learning algorithm. We used a wavelet filter for generating the denoised neural features to train an Artificial Neural Network (ANN) for speech decoding. We found that wavelet based denoising increased the SNR of the neural signal prior to analysis and facilitated accurate speech decoding performance using as few as 40 single-trials. This study may open up the possibility of limiting MEG trials for other task evoked studies as well.

AB - Advancing the knowledge about neural speech mechanisms is critical for developing next-generation, faster brain computer interface to assist in speech communication for the patients with severe neurological conditions (e.g., locked-in syndrome). Among current neuroimaging techniques, Magnetoencephalography (MEG) provides direct representation for the large-scale neural dynamics of underlying cognitive processes based on its optimal spatiotemporal resolution. However, the MEG measured neural signals are smaller in magnitude compared to the background noise and hence, MEG usually suffers from a low signal-to-noise ratio (SNR) at the single-trial level. To overcome this limitation, it is common to record many trials of the same event-task and use the time-locked average signal for analysis, which can be very time consuming. In this study, we investigated the effect of the number of MEG recording trials required for speech decoding using a machine learning algorithm. We used a wavelet filter for generating the denoised neural features to train an Artificial Neural Network (ANN) for speech decoding. We found that wavelet based denoising increased the SNR of the neural signal prior to analysis and facilitated accurate speech decoding performance using as few as 40 single-trials. This study may open up the possibility of limiting MEG trials for other task evoked studies as well.

KW - Artificial neural network

KW - MEG

KW - Speech

KW - Wavelets

UR - http://www.scopus.com/inward/record.url?scp=85058532794&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85058532794&partnerID=8YFLogxK

U2 - 10.1007/978-3-030-05587-5_16

DO - 10.1007/978-3-030-05587-5_16

M3 - Conference contribution

C2 - 31768504

AN - SCOPUS:85058532794

SN - 9783030055868

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 163

EP - 172

BT - Brain Informatics - International Conference, BI 2018, Proceedings

A2 - Yang, Yang

A2 - Yamamoto, Vicky

A2 - Wang, Shouyi

A2 - Jones, Erick

A2 - Su, Jianzhong

A2 - Mitchell, Tom

A2 - Iasemidis, Leon

PB - Springer Verlag

Y2 - 7 December 2018 through 9 December 2018

ER -

Determining the optimal number of MEG trials: A machine learning and speech decoding perspective

Abstract

Publication series

Other

Keywords

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this