[Eeglablist] PCA for Artifact Removal

[Joseph Dien]

Easiest option is to use EP Toolkit:

https://urldefense.com/v3/__https://sourceforge.net/projects/erppcatoolkit/__;!!Mih3wA!X3wDb4yyTgNRFkM50sHhw2Oc1inZY6uJJS-PlSixkcXfPxE1tY4-OUeYNuFExSCB3QhCEQ$ 

Joe

> On Apr 21, 2020, at 16:39, Nick Heredia <n_heredia at u.pacific.edu> wrote:
> 
> Hello all,
> 
> I am trying to remove blink and eye movement artifacts from eeg without the
> help of EOGs. From what I understand, EEGlab (I have 2019.0) has runpca and
> runpca2 <https://sccn.ucsd.edu/pipermail/eeglablist/2018/014193.html>, and
> also pop_runICA has an argument 'pca'
> <https://sccn.ucsd.edu/pipermail/eeglablist/2019/014819.html>.
> 
> My first question is, I explored the Chapters, Plotting ERP Images,
> Decomposing Data Using ICA, Working with ICA components, Time/Freq
> Decomposition, Rejecting Artifacts, and Event Processing, and I wasn't able
> to find help for using PCA in pop_runica. Could you point me in the right
> direction?
> 
> Secondly, my basic understanding is that I can use PCA to remove the
> artifacts for eye movements by removing the highest PCA component
> <https://urldefense.com/v3/__https://www.reddit.com/r/eeg/comments/fk7gve/which_technique_can_be_used_to_remove_these/fkr6yo0?utm_source=share&utm_medium=web2x__;!!Mih3wA!U8PrlZYujMuwwI6LsDEiIgPelDI-xnYcqAJQ5NQG6YZJbxIavIht55xlNX_AW6MxnZqZbg$ >.
> I* have to use PCA by professor's instruction*. My confusion stems from the
> fact that when I use runpca, I get a vector of equal size to my data, (say
> n channels by 24000 samples). I try to interpolate the outlier values in
> this matrix, but I don't know where to go. How would I get the time series
> out of the modified pc output from runpca? Additionally, I tried to redo
> the artifact removal using pca()
> <https://urldefense.com/v3/__https://www.mathworks.com/help/stats/pca.html__;!!Mih3wA!U8PrlZYujMuwwI6LsDEiIgPelDI-xnYcqAJQ5NQG6YZJbxIavIht55xlNX_AW6PL4vlgFQ$ > from the matlab library,
> citing the reddit link.. From the 'PCA Using ALS for Missing Data' section,
> I get a coeff1 matrix of n by n. How would I identify and remove the
> 'highest' PCA component? And would the runpca or pop_runica be able to
> accomplish what I'm trying to do?
> 
> -- 
> With Gratitude
> -N. Heredia
> _______________________________________________
> 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

--------------------------------------------------------------------------------

Joseph Dien, PhD
Senior Research Scientist
Department of Human Development and Quantitative Methodology
University of Maryland, College Park
https://urldefense.com/v3/__http://joedien.com__;!!Mih3wA!X3wDb4yyTgNRFkM50sHhw2Oc1inZY6uJJS-PlSixkcXfPxE1tY4-OUeYNuFExSDl_Xz8Wg$