EEG-ECoG recording

Oosugi Naoya

• Processing

  The data were processed in MatLab.

  • Band-pass filter

    4th order butter worth filter(Signal Processing Toolbox)

    Regression

    PCA+Linear regression(Statistics Toolbox)
    

    PCA was used as whitening.

    Model Estimation

    4-fold cross validation test

    It did NOT destroy the structure of time-series.

    Prediction rate

    Correlation coefficient between test data and predicted data

  Result

• 1. Plotted EEG prediction rate via ECoG on head map

  EEGs and ECoGs were bandpass-filtered between 1 and 45 Hz. Color bar means prediction rate. Black points means locations of EEG channels(without Cz) but these are not correct because this subject is monkey! This figure shows that all EEG channels can be predicted via ECoG and prediction rate of left EEGs is better than right.

• 2. Compared EEG prediction rate via ECoG during time-frequency bands

  EEGs and ECoGs were bandpass-filtered in the time-frequency bands of Theta(between 4 and 7Hz), Alpha(between 8 and 13Hz), low Beta(between 14 and 20Hz), high Beta(between 21 and 30Hz) and Gamma(between 31 and 45Hz). x-axis means time-frequency bands and y-axis means locations of EEG channels. Color bar means prediction rate. This figure shows high frequency components of EEGs are harder to be predicted via ECoGs.

• 3. Prediction rate of EEGs in the frequency band of Theta via ECoGs in the frequency band of Theta or not

  EEGs in the frequency band of Theta were predicted via ECoGs in the frequency band of Theta and ECoGs without the frequency band of Theta(bandcut-filtered between 4 and 7Hz). x-axis means locations of EEG channels and y-axis means prediction rate. Blue bar means prediction rate of Thete EEGs via Theta ECoGs and red bar means prediction rate of Theta EEGs via EEGs without Theta components. This figure shows the time-frequency band of EEG and ECoG is very similar.