I’m quite new to MNE and the more technical side of neuroscience in general, so please forgive if I’m making some obvious mistakes.
I’m working on a project for which I want to do a combined analysis of fnirs and eeg resting data to look in to neurovascular coupling. I have around 70 participants and the data is already down sampled but still too large, I will need to do a dimensionality reduction of the eeg data before doing the combined analysis with the fnirs data for each participant. My questions are:
What method do you think is most suited for this task? In the end I would like to end up with a raw object reduced to a few channels (whatever they will represent will depend on the method of course) that represent as much of the original data as possible for each participant. Methods that I think might be suitable are spatio-temporal clustering and source localization, but please let me know if you have other suggestions.
Do you see any problem if i use this data then after for the combination with the fnirs data and later on for group level statistics?
Hello,
I would really appreciate some feedback! Or if I have asked a stupid / too general question could you maybe give me some feedback on how to improve my original question?
Best regards,
Lennart
I am not too familiar with fNIRS but I know that it measures blood-oxygenation similar to fMRI. It’s a different measure of neural activity than EEG which measure postsynaptic currents. You will have to dig in the literature a bit to understand what methods might be best for appropriate comparison of BOLD activity and EEG/MEG. This might be a good starting point for example: https://journals.physiology.org/doi/full/10.1152/jn.01005.2009
Hello Mainak,
Thanks a lot for your Response! I probably didn’t make it clear, but I am using both EEG and fNIRS data. I am using the mne-nirs package you mentioned for the preprocessing of the fNIRS data. The main Problem I have right now though is that for the method I would like to use (cross frequency coupling between EEG and fNIRS) I need to somehow reduce the dimensionality of the EEG data to reduce the computational resources needed. The package I would like to use for the combined analysis is this one btw: GitHub - pactools/pactools: Phase-amplitude coupling (PAC) toolbox (it interfaces with mne). So my question is: Which method do you think would be most suitable to reduce the eeg data (eg. ICA/PCA, clustering, source estimation etc.) so when I later on do the cross frequency coupling analysis with the fnirs data I for example only have to calculate it for eg. 10 eeg channels instead of the original 32? Sorry again if my wrong utilization of the terminology is confusing, I hope I could get across where my problem lies. Thanks so much again!
Best Regards,
Lennart
