Optimal kernel time-frequency representation of EEG signals in patients with spinal cord injury

H. Lakany

    Research output: Contribution to conferencePaper

    Abstract

    The goal of this study is to develop signal processing methods that can be used to enhance visualization and feature extraction of EEG signals for use in developing brain-computer interfaces (BCIs) for patients with spinal-cord injury (SCI). In this work, a method to represent EEG signals in the time-frequency domain by designing an optimal kernel function which allows for the extraction of maximum possible time-frequency information. The study is based on EEG data recorded from 3 SCI patients attempting to perform 30sec periods of sustained voluntary contractions of the muscles affected by the SCI. Testing protocols were approved by a local Ethics Committee. 30 channels of EEG were recorded together with wrist and ankle muscle EMG when the patient was relaxed and during periods when the patient attempted to generate a sustained wrist extension or ankle dorsi-flexion. Data was collected from each patient on two separate occasions at a minimum interval of 10 weeks. The method we propose here is based on Wigner-Ville distribution which has proved superior to short-time Fourier and Wavelet transforms in visualizing the energy distribution in time-frequency domain. From the time-frequency representation proposed, we extracted information such as moments and marginals and identified the number of elementary sources contributing to the EEG signal during the attempted movement. For each subject, estimated moments and marginals were similar for each test day. This is interpreted as an indication that the extracted signal associated with the performance of an attempted voluntary contraction is relatively stable and can be considered independent of any spatio-temporal reorganization that may have happened in the motor system in response to a spinal lesion in the interval between test sessions. We conclude that this information may potentially be considered as salient features in developing non-invasive BCI applications for SCI patients.

    Conference

    ConferenceSociety for Neuroscience Annual meeting 2006
    CountryUnited States
    CityAtlanta
    Period14/10/06 → …

    Fingerprint

    Spinal Cord Injuries
    Electroencephalography
    Brain-Computer Interfaces
    Wrist
    Ankle
    Wavelet Analysis
    Ethics Committees
    Fourier Analysis
    Skeletal Muscle
    Muscles

    Keywords

    • signal processing
    • visualization
    • brain-computer interfaces

    Cite this

    Lakany, H. (2006). Optimal kernel time-frequency representation of EEG signals in patients with spinal cord injury. Paper presented at Society for Neuroscience Annual meeting 2006, Atlanta, United States.
    Lakany, H. / Optimal kernel time-frequency representation of EEG signals in patients with spinal cord injury. Paper presented at Society for Neuroscience Annual meeting 2006, Atlanta, United States.
    @conference{c70f62735fe04a43ae50e4a478a07e4d,
    title = "Optimal kernel time-frequency representation of EEG signals in patients with spinal cord injury",
    abstract = "The goal of this study is to develop signal processing methods that can be used to enhance visualization and feature extraction of EEG signals for use in developing brain-computer interfaces (BCIs) for patients with spinal-cord injury (SCI). In this work, a method to represent EEG signals in the time-frequency domain by designing an optimal kernel function which allows for the extraction of maximum possible time-frequency information. The study is based on EEG data recorded from 3 SCI patients attempting to perform 30sec periods of sustained voluntary contractions of the muscles affected by the SCI. Testing protocols were approved by a local Ethics Committee. 30 channels of EEG were recorded together with wrist and ankle muscle EMG when the patient was relaxed and during periods when the patient attempted to generate a sustained wrist extension or ankle dorsi-flexion. Data was collected from each patient on two separate occasions at a minimum interval of 10 weeks. The method we propose here is based on Wigner-Ville distribution which has proved superior to short-time Fourier and Wavelet transforms in visualizing the energy distribution in time-frequency domain. From the time-frequency representation proposed, we extracted information such as moments and marginals and identified the number of elementary sources contributing to the EEG signal during the attempted movement. For each subject, estimated moments and marginals were similar for each test day. This is interpreted as an indication that the extracted signal associated with the performance of an attempted voluntary contraction is relatively stable and can be considered independent of any spatio-temporal reorganization that may have happened in the motor system in response to a spinal lesion in the interval between test sessions. We conclude that this information may potentially be considered as salient features in developing non-invasive BCI applications for SCI patients.",
    keywords = "signal processing , visualization, brain-computer interfaces",
    author = "H. Lakany",
    year = "2006",
    language = "English",
    note = "Society for Neuroscience Annual meeting 2006 ; Conference date: 14-10-2006",

    }

    Lakany, H 2006, 'Optimal kernel time-frequency representation of EEG signals in patients with spinal cord injury' Paper presented at Society for Neuroscience Annual meeting 2006, Atlanta, United States, 14/10/06, .

    Optimal kernel time-frequency representation of EEG signals in patients with spinal cord injury. / Lakany, H.

    2006. Paper presented at Society for Neuroscience Annual meeting 2006, Atlanta, United States.

    Research output: Contribution to conferencePaper

    TY - CONF

    T1 - Optimal kernel time-frequency representation of EEG signals in patients with spinal cord injury

    AU - Lakany, H.

    PY - 2006

    Y1 - 2006

    N2 - The goal of this study is to develop signal processing methods that can be used to enhance visualization and feature extraction of EEG signals for use in developing brain-computer interfaces (BCIs) for patients with spinal-cord injury (SCI). In this work, a method to represent EEG signals in the time-frequency domain by designing an optimal kernel function which allows for the extraction of maximum possible time-frequency information. The study is based on EEG data recorded from 3 SCI patients attempting to perform 30sec periods of sustained voluntary contractions of the muscles affected by the SCI. Testing protocols were approved by a local Ethics Committee. 30 channels of EEG were recorded together with wrist and ankle muscle EMG when the patient was relaxed and during periods when the patient attempted to generate a sustained wrist extension or ankle dorsi-flexion. Data was collected from each patient on two separate occasions at a minimum interval of 10 weeks. The method we propose here is based on Wigner-Ville distribution which has proved superior to short-time Fourier and Wavelet transforms in visualizing the energy distribution in time-frequency domain. From the time-frequency representation proposed, we extracted information such as moments and marginals and identified the number of elementary sources contributing to the EEG signal during the attempted movement. For each subject, estimated moments and marginals were similar for each test day. This is interpreted as an indication that the extracted signal associated with the performance of an attempted voluntary contraction is relatively stable and can be considered independent of any spatio-temporal reorganization that may have happened in the motor system in response to a spinal lesion in the interval between test sessions. We conclude that this information may potentially be considered as salient features in developing non-invasive BCI applications for SCI patients.

    AB - The goal of this study is to develop signal processing methods that can be used to enhance visualization and feature extraction of EEG signals for use in developing brain-computer interfaces (BCIs) for patients with spinal-cord injury (SCI). In this work, a method to represent EEG signals in the time-frequency domain by designing an optimal kernel function which allows for the extraction of maximum possible time-frequency information. The study is based on EEG data recorded from 3 SCI patients attempting to perform 30sec periods of sustained voluntary contractions of the muscles affected by the SCI. Testing protocols were approved by a local Ethics Committee. 30 channels of EEG were recorded together with wrist and ankle muscle EMG when the patient was relaxed and during periods when the patient attempted to generate a sustained wrist extension or ankle dorsi-flexion. Data was collected from each patient on two separate occasions at a minimum interval of 10 weeks. The method we propose here is based on Wigner-Ville distribution which has proved superior to short-time Fourier and Wavelet transforms in visualizing the energy distribution in time-frequency domain. From the time-frequency representation proposed, we extracted information such as moments and marginals and identified the number of elementary sources contributing to the EEG signal during the attempted movement. For each subject, estimated moments and marginals were similar for each test day. This is interpreted as an indication that the extracted signal associated with the performance of an attempted voluntary contraction is relatively stable and can be considered independent of any spatio-temporal reorganization that may have happened in the motor system in response to a spinal lesion in the interval between test sessions. We conclude that this information may potentially be considered as salient features in developing non-invasive BCI applications for SCI patients.

    KW - signal processing

    KW - visualization

    KW - brain-computer interfaces

    UR - http://www.sfn.org/index.aspx?pagename=neuroscienceQuarterly_06Summer_neuroscience2006

    UR - http://www.sfn.org/index.aspx?pagename=abstracts_ampublications

    M3 - Paper

    ER -

    Lakany H. Optimal kernel time-frequency representation of EEG signals in patients with spinal cord injury. 2006. Paper presented at Society for Neuroscience Annual meeting 2006, Atlanta, United States.