Source analysis in frequency domain

Discussions
1

Hi,
I’m trying to analysis about human’s fatigue state classification using source map features.
Brainstorm tool is very cool and easy to use the MATLAB, so I will try to analysis using this toolbox.

So, my question is how can we extract the frequency domain feature such as power spectral density in source space. I want to know how alpha, beta, and theta PSD values change in source space. But I haven’t seen it discussed before.

EEG signal -> Source map projection -> FFT
or EEG signal -> FFT -> Source map projection
I don’t know what is the right analysis approach…

I think this toolbox give the first approach, but this requires a lot of memory and computation, right?
So, in my opinion, apply the EEG signal -> FFT in MATLAB, and import these FFT mat file in Brainstorm for source map result is the best.

But, I’m not sure this approach is the right method.
So, I want to know about this approach is right TT

Thanks,

Raehyun

2

You won’t be able to estimate the source activity from frequency power: the electromagnetic models require the EEG signals with positive and negative values.

The recommended approach is indeed to estimate the sources and then compute the power or time-frequency decompositions. For example: https://neuroimage.usc.edu/brainstorm/Tutorials/RestingOmega
To reduce the amount of memory needed: use only a few frequency bands, or a few regions of interest.

Another valid approach is possible, more along the lines of what you want to do: filter all your recordings (task data and noise/rest recordings) in a given frequency band at the beginning of your analysis. Then go through all the recommended processing pipeline (cleaning, epoching, source reconstruction)

3

Thank you very much !!! :slight_smile: