Changes in EEGlab/fmrib versions

[arnodelorme]

Copied from https://sccn.ucsd.edu/bugzilla/show_bug.cgi?id=1079

This bug might not be relevant any more since it refers to an old version of fMRIb

Hello,

I tried the analyses with the 4 different combinations of version, and found the same results with:
- eeglab6.01b + fmrib1.2 and eeglab9_0_4_6s + fmrib1.2
- eeglab6.01b + fmrib1.21 and eeglab9_0_4_6s + fmrib1.21

My first result with eeglab6.01b + fmrib1.2 (presented on the figure as 01-V6) is weird as I did not manage to reproduce it...

So it seems that the new version of the fmrib plugin does not correct the artefacts in the same way than the old one.

I have attached the batch I use to run the correction, but the data is unfortunately too big (>1G).

Let me no if I can do anything else,
Best,
katia
[reply] [−]Comment 4Lehongre 2011-07-04 03:04:27 PDT
Created attachment 561 [details]
batch
[reply] [−]Comment 3Arnaud Delorme 2011-07-01 13:25:28 PDT
Hi Katia,

could you upload the dataset you have used if it is not too big. Also, what about using eeglab6.01b + fmrib1.21. Does it give the same result as eeglab6.01b + fmrib1.2. If it does then it makes the problem simpler because it means that the problem is with EEGLAB and not with fmrib1.21.

Best,

Arno
[reply] [−]Comment 2Lehongre 2011-06-14 03:32:11 PDT
Just in case, here is the command line I've used for the artefact correction:

EEG = pop_fmrib_fastr(EEG,125,4,40,'VS',0,0,0,0,0,0.03,[31 32],'auto');

katia
[reply] [−]Comment 1Lehongre 2011-06-12 03:36:17 PDT
Created attachment 556 [details]
Exemple of resulting corrections
[reply] [−]DescriptionLehongre 2011-06-12 03:35:20 PDT
Hi,

I've reanalyzed two years old EEG-fMRI data with the upgraded versions of EEGlab and fmrib plugin, and did not find the same artefact correction.
More precisely the first time I've used the eeglab6.01b and fmrib1.2 versions, and now I'm using eeglab9_0_4_6s and fmrib1.21.

To find out the source of the problem, I ran in parallel the same analysis on the same dataset, on two different computers, each one with a different eeglab/fmrib version but with identical Matlab version.
I observed that the number of residual PCs removed by channel is completely different from one version to another. For a given channel, it happened to be 1 for the new version and 9 for the old one, or reversed, but globally the new version seemed to remove less PCs.

I also tested an analysis with the new eeglab version and the old fmrib version, and found a third type of correction!
In the jpg file attached, I've plotted the resulting corrections for channel O1 with :
- eeglab6.01b + fmrib1.2 (O1-V6)
- eeglab9_0_4_6s + fmrib1.21 (O1-V9)
- eeglab9_0_4_6s + fmrib1.2 (O1-V9-fmrib12)

Do you have any idea of what has changed? Which correction should I use? The impact on the final analyses (time frequencies) is unfortunately quite important.

Best regards,
katia