[FieldTrip] Fwd: Source localisation mom results

Thu Apr 23 07:56:55 CEST 2020

Dear Xavier,

Thanks for your message. Let me forward your questions to the fieldtrip discussion list. If you haven’t signed up yet, please do so (http://www.fieldtriptoolbox.org/discussion_list/).

I prefer to have the interaction through the list, because then more people will benefit from the discussion. Also, other people who have an opinion about the questions can also chime in. Both points make the transfer of knowledge more efficient.

At the moment, I don’t have much time to look in detail at your questions, since we are running our annual toolkit course at the Donders Centre right now (this year for the first time fully online). Therefore, I hope that somebody reading this might pick up on your message.

Some quick pointers:
1.  It depends on the connectivity metric of interest, but for most I’d compute the source level data for the frequency filtered signals (assuming a beamformer). When you’d be using a distributed source modelling approach (e.g. minimum norm estimate) the order probably doesn’t matter too much.
2. No, 2 different measurements are likely too different (different electrode positions, SNR characteristics etc), so that appending them before the source reconstruction does not make sense.
3. I did not check in detail, but your descriptions looks OK. This would be one way to get ‘parcellated’ results.
4. In it’s ’naive’ (i.e. without additional settings) application the beamformer has a well-known ‘depth bias’, resulting in a large dynamic range of the amplitude of the dipoles, with dipole locations that are far away from the sensors getting very large amplitudes. See http://www.fieldtriptoolbox.org/tutorial/beamformer/ for another example (although it’s MEG, the same thing applies).

Best wishes,
Jan-Mathijs

Begin forwarded message:

From: Xavier Vrijdag <x.vrijdag at auckland.ac.nz<mailto:x.vrijdag at auckland.ac.nz>>
Subject: Source localisation mom results
Date: 23 April 2020 at 02:12:02 CEST
To: "jan.schoffelen at donders.ru.nl<mailto:jan.schoffelen at donders.ru.nl>" <jan.schoffelen at donders.ru.nl<mailto:jan.schoffelen at donders.ru.nl>>

Hello Jan-Mathijs,

I have been using fieldtrip for my EEG analysis for the last year and am really pleased with all the options the toolbox offers and the support given along my journey to understand how to use the toolbox. The documentation on the website is really helpful to understand how to use the various parts of the toolbox.

So far I have programmed a preprocessing pipeline to clean my EEG data that I collected during exposures to various narcotic gases (resting state data). This data is fed into a source localization script with the ultimate goal to do connectivity analysis (mutual information) and network analysis.

I have the following questions:

  1.  Should I filter the EEG data into the standard frequency bands before the source localization, or is it better to filter the mom data that is fed into the connectivity analysis. With the goal to assess the connectivity in various frequency bands
  2.  Should I append the data from two measurements (separate days)  for each participants or keep them separate during the source localization analysis (each measurement contains a baseline and exposure recording).
  3.  Did I correctly use the parcellation functions to source-localize the data to the centroids of the 90 cortical regions?
  4.  I intend to feed the mom field of the parcellated source struct into the connectivity analysis. I found that the range of the is 6.0571e+07. Is this correct/normal?

The script has the following steps (script file can be downloaded from:https://www.dropbox.com/s/jnf5dr56ytj543j/EEGsourselocalisationN2He.m?dl=0):
I currently have a source localization script  that has the following steps:
I have a created a file with the aligned electrode locations
I use the standard BEM head model and the AAL atlas

create a 6 mm grid based leadfield
source interpolate the AAL atlas on the grid
calculate the centroid position for each of the first 90 AAL regions
create new grid based on these center positions
new source interpolation of the AAL atlas on the new grid

Append the baseline and exposure data
Frequency band filter
Calculate the channel covariance matrix for all exposures and for each exposure separately (keep trials)
Calculate LCMV beamformer spatial filter on all exposure matrix
Apply spatial filter and source parcellation on each exposure (keep trials and rawtrial)

Thank you for helping me getting a better understanding of the source localization steps.

Regards,

Xavier Vrijdag

Xavier Vrijdag, MSc

Researcher
Department of Anaesthesiology │ School of Medicine
Faculty of Medical & Health Sciences │ The University of Auckland
Private Bag 92019 │ Auckland 1142 │ New Zealand

M +64 21 0230 4558
E x.vrijdag at auckland.ac.nz<mailto:x.vrijdag at auckland.ac.nz>

[FMHS Logo]

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20200423/c99901b7/attachment.htm>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: image001.png
Type: image/png
Size: 22256 bytes
Desc: image001.png
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20200423/c99901b7/attachment.png>