[Eeglablist] Help with EEG data analysis and power band feature extraction.

[Velu Prabhakar Kumaravel]

Hi Jorge,

Regarding the preprocessing pipeline, I do not know the Sampling Rate and
the number of EEG channels, but 3 minutes of data is probably insufficient
for a good ICA decomposition. In this case, a PCA is recommended
<https://urldefense.com/v3/__https://eeglab.org/tutorials/06_RejectArtifacts/RunICA.html*which-ica-algorithm__;Iw!!Mih3wA!Aoqj02vNGOH5p55o7YzahzBYrF1fxoV9rUG7NfyVzy5isWLiY6zwkbu1NbM54y_db4GZwKlmsv4dQrNCByCPrxiXxC_aoqn9$  >before
ICA (or you could clean data using ASR before ICA). Also, your high-pass
filter setting might not be optimal for ICA. It is recommended
<https://urldefense.com/v3/__https://eeglab.org/tutorials/06_RejectArtifacts/RunICA.html*how-to-deal-with-the-aggressive-high-pass-filter-applied-before-running-ica__;Iw!!Mih3wA!Aoqj02vNGOH5p55o7YzahzBYrF1fxoV9rUG7NfyVzy5isWLiY6zwkbu1NbM54y_db4GZwKlmsv4dQrNCByCPrxiXxEVVbaJe$  >to
use at least 1 Hz high-pass.

Also, I'd remove bad channels before performing the ICA. When the data is
relatively cleaner
<https://urldefense.com/v3/__https://eeglab.org/tutorials/06_RejectArtifacts/RunICA.html*what-do-we-mean-by-obtaining-a-better-ica-decomposition__;Iw!!Mih3wA!Aoqj02vNGOH5p55o7YzahzBYrF1fxoV9rUG7NfyVzy5isWLiY6zwkbu1NbM54y_db4GZwKlmsv4dQrNCByCPrxiXxHnPHZTE$  >,
ICA can be more precise.

Best,

Velu Prabhakar Kumaravel, PhD Student
Center for Mind/Brain Sciences,
University of Trento, Italy


On Tue, 25 Oct 2022 at 16:13, Jorge Delgado-Muñoz <
delgadojorge at g.ecc.u-tokyo.ac.jp> wrote:

> Dear all.
>
> I am collecting EEG and PPG data simultaneously  from volunteers while
> performing a memory span task, in order to find correlation between EEG
> rhythms (alpha, beta and theta)  and Heart Rate variability.
>
> However, due to my lack of experience in processing EEG continuous data, I
> am facing some difficulties in order to achieve this. For now, my pipeline
> to process the EEG data I am collecting goes as follows.
>
> 1. Import raw data collected from g.tec gUSBamp via Matlab into EEGLAB .
> set format.
> 2. Bandpass filter data (0.5hz to 30Hz).
> 3. Remove eye blinking artifacts using ICA decomposition.
> 4. Interpolate bad channels (in some recordings, I have had to interpolate
> up to three channels). The criteria to consider bad channels is performing
> visualization of data stream, spectra and results of cleaning raw data and
> ASR.
>
> This procedure is performed on every EEG continuous recording individually,
> whose length varies between 3 and 8 minutes. What I did before was to use
> the function STUDY of EEGLAB GUI and compare the power frequency spectra of
> the whole recording sets using parametric statistics using the condition
> (rest, task) as an independent variable.
>
> However, I consider this method a bit naive and I would like to improve it,
> and to do so, I would like to check what is wrong with my preprocessing
> pipeline, especially regarding the use of ICA and channel interpolation.
> Also, I am analyzing the HRV data breaking down every recording into
> 1-minute segments, from which I can extract a number of features. My
> question is, should I use the same temporality with EEG?, that means, to
> divide each dataset into the corresponding 1-minute segment synchronized
> with HRV and process each segment individually. Also, how can I, for
> example, extract  numerical values or features related to frequency power,
> so I can use it to perform correlation or other statistical analysis on
> both signals?
>
>
> Thank you very much for your attention, I hope I can refine my method so I
> can perform an adequate analysis.
>
>
>
> Best Regards,
>
> --
>
> Jorge Delgado-Muñoz
>
> Hiraki Lab
>
>
> https://urldefense.com/v3/__https://ardbeg.c.u-tokyo.ac.jp/en/top-2/__;!!Mih3wA!DmzpQuAnwTVTpMfOkkxlBo3e5Ouorz2iVnAxsgJTzicDtMLV7V3vkehg-ABwTyMVZSVjMYBctTiMgximFfC8SOsPstbNmNxiJ40NKtY$
>
> <
> https://urldefense.com/v3/__https://eur03.safelinks.protection.outlook.com/?url=https*3A*2F*2Fardbeg.c.u-tokyo.ac.jp*2Fen*2Ftop-2*2F&data=04*7C01*7CYutaro.Hamasaki*40tobii.co.jp*7Ca7f9c067c478445bffd408d874055269*7C6b7dddeaad8f4522be45cb8e69a29a5a*7C0*7C0*7C637386913512066949*7CUnknown*7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0*3D*7C1000&sdata=qi4VdnXRljHwyiC0q5NyRgG5QgdI4ggqzFM2g8pe5*2Bw*3D&reserved=0__;JSUlJSUlJSUlJSUlJSUlJSUlJSU!!Mih3wA!DmzpQuAnwTVTpMfOkkxlBo3e5Ouorz2iVnAxsgJTzicDtMLV7V3vkehg-ABwTyMVZSVjMYBctTiMgximFfC8SOsPstbNmNxiMAdxq7g$
> >
>
> Department of General System Studies
>
> Graduate School of Art and Science, The University of Tokyo
> email:  delgadojorge at g.ecc.u-tokyo.ac.jp
>             joardemu at gmail.com
> _______________________________________________
> Eeglablist page: http://sccn.ucsd.edu/eeglab/eeglabmail.html
> To unsubscribe, send an empty email to
> eeglablist-unsubscribe at sccn.ucsd.edu
> For digest mode, send an email with the subject "set digest mime" to
> eeglablist-request at sccn.ucsd.edu