misc concerns and questions with mne_analyze and 256-channel EEG data

✉️ Mailing List Archive (read-only)
1

Hi MNE analyzers,

I have just succeeded in manipulating my data so that I can use mne_analyze for interactive analysis. So far I have

Exported EEG as EDF file
Converted to FIF file
Aligned electrodes to scalp surface
Created a single event and averaged so I could treat my raw data as averaged data (only way to get to into mne_analyze, I am told)
Created BEM, forward, and inverse
Loaded all into mne_analyze

Questions & concerns:

Source solutions on inflated surface do not seem to bear any relation to the scalp EEG data plotted on the scalp surface
I cannot get the MRI solutions to change as I change the time sample, thought the other displays behave appropriately
Also, the single MRI solution looks like the dMNE ones, and does not change as I switch to MNE or sLORETA estimates
I cannot figure out how to display any EEG on the right-hand Data window, even after choosing "Adjust/select trace layout/sample EEG"
I can find not mention in the manual on how to adjust the display amplitude of the source waveforms in the bottom left pane
MNE source locations are fairly evenly spread out across both hemispheres, but dSMP and sLORETA's are mainly confined to R hemi
All the source waveforms appear to have DC shifts lasting for the entire epoch ? this does not seem right

All in all, I am not comfortable with what I am observing, and wonder if I have dome something grossly wrong in my analysis stream so far. I am post files and screenshots to DropBox if anyone is willing to assist me with any of the above.

The best thing I can think of doing at this point is to create some simulated time courses and EEG and run it through the steps I have performed above.

-Jeff Eriksen

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.nmr.mgh.harvard.edu/pipermail/mne_analysis/attachments/20120608/4df967b0/attachment.html

2

Hi see follow-ups in text

Hi MNE analyzers,

I have just succeeded in manipulating my data so that I can use mne_analyze for interactive analysis. So far I have

Exported EEG as EDF file
Converted to FIF file
Aligned electrodes to scalp surface
Created a single event and averaged so I could treat my raw data as averaged data (only way to get to into mne_analyze, I am told)
Created BEM, forward, and inverse
Loaded all into mne_analyze

Questions & concerns:

Source solutions on inflated surface do not seem to bear any relation to the scalp EEG data plotted on the scalp surface

What does this mean? Could you explain with more detail? Possibly send a screenshot? (though see below, it sounds like there may be some issues in earlier stages, which would be better to send first)

I cannot get the MRI solutions to change as I change the time sample, thought the other displays behave appropriately

What do you mean by MRI solutions?

Also, the single MRI solution looks like the dMNE ones, and does not change as I switch to MNE or sLORETA estimates

I'm not sure what you mean by "single MRI solutions" or "dMNE"? What windows are you talking about please describe more details.

I cannot figure out how to display any EEG on the right-hand Data window, even after choosing "Adjust/select trace layout/sample EEG"

Did you generate an EEG layout? See the manual ( mne_make_eeg_layout )

I can find not mention in the manual on how to adjust the display amplitude of the source waveforms in the bottom left pane
MNE source locations are fairly evenly spread out across both hemispheres, but dSMP and sLORETA's are mainly confined to R hemi
All the source waveforms appear to have DC shifts lasting for the entire epoch ? this does not seem right

Well, if there are these long DC shifts there should be evidence in the evoked data (or raw if this is raw data). Did you use baselines e.g. bmin bmax? How clean are the averages? (maybe send a plot) How clean is your noise covariance data (did you go through the raw data used to calculate it).

D

3

Dan, see my answers below,
-Jeff

Hi see follow-ups in text

Hi MNE analyzers,

I have just succeeded in manipulating my data so that I can use
mne_analyze for interactive analysis. So far I have

Exported EEG as EDF file
Converted to FIF file
Aligned electrodes to scalp surface
Created a single event and averaged so I could treat my raw data as
averaged data (only way to get to into mne_analyze, I am told)
Created BEM, forward, and inverse
Loaded all into mne_analyze

Questions & concerns:

Source solutions on inflated surface do not seem to bear any relation
to the scalp EEG data plotted on the scalp surface

What does this mean? Could you explain with more detail? Possibly send a
screenshot? (though see below, it sounds like there may be some issues in
earlier stages, which would be better to send first)

I have posted my files and some screenshots to DropBox:
https://dl.dropbox.com/u/74096906/MNE/MNE_files.zip
https://dl.dropbox.com/u/74096906/MNE/MNE_pix.zip
For the time sample I have shown on the pictures, the sources should be
concentrated in the frontal midline based on the scalp topography, but the
sources on the inflated brain surface shows widespread activity with no
obvious concentration in the fontal midline.

I cannot get the MRI solutions to change as I change the time sample,
thought the other displays behave appropriately

What do you mean by MRI solutions?

I mean the source activity shown on the 3D MRI display - subject's own MRI

Also, the single MRI solution looks like the dMNE ones, and does not
change as I switch to MNE or sLORETA estimates

I'm not sure what you mean by "single MRI solutions" or "dMNE"? What
windows are you talking about please describe more details.

I mean the color-coded activity shown in the 3D MRI display for a single
time sample. By dMNE I actually meant the MNE estimate, one of three
available to display, the other two being dSPM and sLORETA.

I cannot figure out how to display any EEG on the right-hand Data
window, even after choosing "Adjust/select trace layout/sample EEG"

Did you generate an EEG layout? See the manual ( mne_make_eeg_layout )

I will look at that again. I assumed it would be able to find and use the
set of layouts I made for the raw EEG.

I can find no mention in the manual on how to adjust the display
amplitude of the source waveforms in the bottom left pane
MNE source locations are fairly evenly spread out across both
hemispheres, but dSMP and sLORETA's are mainly confined to R hemi
All the source waveforms appear to have DC shifts lasting for the
entire epoch ? this does not seem right

Well, if there are these long DC shifts there should be evidence in the
evoked data (or raw if this is raw data). Did you use baselines e.g. bmin
bmax? How clean are the averages? (maybe send a plot) How clean is your
noise covariance data (did you go through the raw data used to calculate
it).

I did not truly average - I faked it out by setting a single event and an
N of 1, so it would seem I had averaged - I was advised to do this so I
could work in mne_analyze and have all the tools available to me that do
not necessarily work for raw data
There _should_ not be any baseline shifts in the raw/averaged data as I
filtered 1-100 Hz, which should have removed the baseline globally for
each channel - thus I did not use bmin/bmax
I calculated the noise-covariance using the entire epoch

4

Hello Jeff,

you should use ico5 or oct6 for the source space. ico3 is too small.
you can try with a diagonal noise covariance if the topography do not
match the source estimates.

Alex

5

One problem here was that a flat channel was included in the inverse calculation. Since its noise is zero, the inverse calculation goes haywire.

- Matti