CODEWITHSUNDEEP
matlabsignal-processing
hussein hammoud
hussein hammoud

Reputation: 385

ICA EEG signal Matlab clarification

I just have a couple of questions about Independent component analysis (ICA) for EEG signals in MATLAB.
I have an EEG data, which consists of 29 channels, and 3600 sec each.
After reading some papers about EEG data processing, I get the idea that I need to do ICA in order to remove the noise from the signal, and stick with only the interesting data which I need to process.
I was trying now to use fastICA which seemed pretty simple and hopefully robust.
Here comes the questions:

  1. about fastICA, I should input my data in the format of a matrix where the # of rows = # of channels, and # of cols = # of seconds right?

  2. If the above is correct, when I run the algorithm, I'm getting a 28-row matrix. Does that mean I have 28 independent sources in my signal?

    1. If I have this result, How I'm supposed to know which independent component corresponds to which source? For example, How I'm going to tell that component 1 corresponds to the eye movement, while component 2 corresponds to another?

These questions came to me while I was analysing my EEG signal, and I couldn't get answers online.
I just want to get some answers, where as other research paper go deep in math, for which I don't have a very good basis to rely on.
Thanks alot for any help, Regards

Upvotes: 0

Views: 2302

Answers (1)

Setsu
Setsu

Reputation: 1218

I work for a lab that publishes EEG data focusing on the cognitive functionality of schizophrenics so I thought I'd chime in. We use EEGLAB to do all our processing and one of the steps we do is ICA to strip out noise from the signal, such as power line noise and eye blinks. Syntax wise it's very easy to use, just a call to pop_runica will do the trick. However, like the other commenters have stressed it is not wise to blindly follow this technique without a good understanding of why your are doing it. I suggest another paper by McMenamin et al. (Validation of ICA-Based Myogenic Artifact Correction for Scalp and Source-Localized EEG); the included supplementary PDF includes examples of ICs of each category with explanations of what they are.

If you're in a research group ask your colleagues for how they use ICA. If that's not an option, you need to dig through the literature some more to convince yourself the validity of this technique. Also, I would recommend a discussion forum that's focused on EEG, such as the EEGLAB mailing list or the FSL support forum.

Upvotes: 2

Related Questions