Source Localization - Dipole Fit in Brainstrom

Discussions
1

Dear Experts,

From the BST forum history we learned that dipole fitting procedures used to be included in BST. Since you spent a lot of effort in implementing algorithms to perform inverse solutions, we would like to ask for your support in this respect.

Aiming a decent source analysis of epilepticform activity, we are planning to perform advanced dipole fitting procedures including fitting of moving dipoles. For this task we would appreciate if you could provide helpful scripts and/or information to us.

We are looking forward to hear from you.

Thank you very much in advance.

Best regards,
Alexander Hunold

2

Dear Alexander:

We are in the process of revitalizing the dipole-fitting processes in Brainstorm and this will include moving dipoles. We shall be able to deliver some elements by September this year.

Best wishes,

3

Dear Sylvain:

Thanks for your response. Those are bright prospects. Mean while we will use the import function of dipoles from *.bdip files as a practical compromise.

Best whishes,
Alexander

4

Dear Experts,

I’m just starting to use BST, I would like to localize the sources after stimulations, so I should localize some sources in the sensitive cortex but it’s nut in the cortex. I’ve import the MRI, the surfaces, the EEG the electrode position but the localization is not very well.
Any ideas about it ?

Thank you by advance.
Gael

5

Hi Gael,

I’m not sure I understand what you are trying to do. Can you please describe:

  • your objectives
  • your methodology
  • what your are trying to get from Brainstorm (attach figures if possible)
  • what you obtain instead of the expected results (attach screenshots of your results)

Thanks
Francois

6

Dear Francois,

Sorry for the delay with wich I reply but I was trying to improve my methodology.
To recap my aim:

  • Objective : I would like to localize the area which are responsible of the epilepsy in order to compare the results with an other software.
  • Methodology : I have several EEG recordings of several patients but I use one of them and I focus on the intercritical spikes to localize the sources.

QUESTIONS:

  • I would like to know how can I choose the best time windows for the noise covariance
  • What is the best methodology to localize properly
    I have already read all informations in the BST tutorials but it’s not enough to have a really hand on it.

Thank you by advance.
Best regards.
Gael

7

Dear experts

I have two questions:I am using colin27 as my default anatomy for EEG source analysis.
1- I would like to have equivalent MNI coordinate of Auditory cortex’s talairach coordinate. How can I transform it?
2- I do not have individual MRI of subjects. How accurate is it to use colin27 as default anatomy.

all the best
Ayda

8

Hello Ayda:

  1. We will be integrating in Brainstorm convenient tools and features to convert coordinates from one referential to the other. For now, I suggest you use tools available online such as: http://www.bioimagesuite.org/Mni2Tal/index.html

  2. Colin27 is just one arbitrary, individual brain: it is not a statistical average from a group: Its size and shape are not representative of the average of the population for instance. You may use it, with various degrees of sophistication, when no individual MRI data is available from your subjects:

a) If you the actual locations of the EEG electrodes on each individual scalp is available, Brainstorm can warp the Colin27 template to the shape and size of the scalp of each participant: it is the most reasonable approximation you can achieve, without individual MRI data. (see http://neuroimage.usc.edu/brainstorm/Tutorials/TutWarping?highlight=(warp))

b) You can project the electrode locations a generic montage onto Colin27’s scalp. Right-click on the channel file > MRI registration > Edit…
If the electrodes are not all clearly on the scalp, use the buttons in the toolbar to align them on the scalp, and then project them on the surface (read the tooltips of the buttons for help)

Solution (a) is favored but requires you have digitized the individual electrode locations.

For the source model per se, to not constrain the source orientations to the cortical surface is arguably a reasonable approach. It is an option available in the inverse model GUI in Expert Mode (http://neuroimage.usc.edu/brainstorm/Tutorials/TutSourceEstimation?highlight=(orientation)).

Hope this helps.

Best wishes,

9

Hello Gael:

  • for noise covariance in case of EEG epileptic data, it is always a bit tricky because there is arguably no segment in the data where the brain activity is not epileptic or actually, just noise. I would suggest for now that you select a fairly long segment of data (the longer in time, the better, but at least 3 times as long as your number of electrodes) where no or little spike activity is visible.

  • the best methodology to localize activity form spikes is still an open debate amongst epileptologists. Some will argue the peak of the spike is the most representation, some will tell that the rising front of the spike is more relevant because afterwards, the epileptic activity has spread out away from the epileptogenic zone. You will need to discuss a strategy with your clinical collaborator/supervisor. My personal choice would be to epoch your data about each spike, in a time window of at least -500ms and +200ms about the spike maximum. Then, run minimum-norm source model on it. Average the spike data and look at its source model and also look at each spike individually to understand the variability in the patterns. Of course, this method can be much refined but this is a start.

Hope this helps.

10

Hello Sylvain,

Thank you very much for your reply and of course that it’s help me !

[I]“My personal choice would be to epoch your data”[/I]
How can I epoch the data in BST ?

[I]“Average the spike data and look at its source model and also look at each spike individually to understand the variability in the patterns”[/I]
Is it possible to average sevral spike diretly in BST ? if yes how can I do that ? I suppose I need to epoch the data before ?

Best regards.
Gael

11

Hi Gael,

Reviewing the recordings, marking events manually, epoching and averaging are topics covered by the following online tutorials:
http://neuroimage.usc.edu/brainstorm/Tutorials/TutRawViewer
http://neuroimage.usc.edu/brainstorm/Tutorials/TutMindNeuromag

Cheers,
Francois

12

Hi Francois,

Thank you for your helpi I will take a look at the tutorial as soon as possible :wink:

Best regards
Gael

13

Hi Francois,

Your last post :

[QUOTE=Francois;3853] Reviewing the recordings, marking events manually, epoching and averaging are topics covered by the following online tutorials:
http://neuroimage.usc.edu/brainstorm/Tutorials/TutRawViewer
http://neuroimage.usc.edu/brainstorm/Tutorials/TutMindNeuromag [/QUOTE]

Yes the tutorial explain how do some task on the data but most of the time just on the MEG data an noway on the EEG data. I’m trying to localize some Evoqued Potential with an averaging of several (around 500) stimulations.
Is it possible to do it ? Is it on the tutorial ?

Best regards.
Gael.

14

Hi Gael,

I’m sorry there is no EEG tutorial yet in Brainstorm.
But yes, you can process EEG recordings with Brainstorm, it works exactly in the same way.
Except that you need to import the position of the electrodes manually, or use one of the defaults available EEG caps in Brainstorm (right-click on a condition > Use default EEG cap)

Francois

15

Hi Francois,

Yes I can process with the EEG recording. Currently I have imported the MRI (.img), the surfaces (.mesh) generated by brainVISA, the EEG recording and the position of the electrodes (ASCII).

Problem :
My EEG recordings and the electrode file don't match. In fact the problem is the channels recording order (EEG) and the order of the electrode localization (Polhemus) aren't the same. The result in BST is that before the electrodes importation and after the EEG file is not the same. The channels of the EEG are modified. They don't match with the real electrodes recording.

Solution :
I solve the problem by a montage generate by an other software exactly in the same order which I have localize the electrodes

Question :
1°) What is the best software to segmentate the T1 MRI ?
2°) Is it possible to average several events when I'm using the EEG recording in order to localize the activated area in response to the evoqued potentials ?
3°) what is the best way to have a good noise covariance :
3.1°) Full VS diagonal matrix
3.2°) What is the best time windows to apply the noise covariance and what is the best conditions (quiet patient, closed eyes, sleeping patient)

My personal choice would be to epoch your data about each spike, in a time window of at least -500ms and +200ms about the spike maximum. Then, run minimum-norm source model on it. Average the spike data and look at its source model and also look at each spike individually to understand the variability in the patterns

4°) How can I epoch my EEG data ?

Thank you in advance.
Best regards.
Gael.

16

Question :
1°) What is the best software to segmentate the T1 MRI ?

We recommend Brainvisa (http://brainvisa.info), Brainsuite (http://neuroimage.usc.edu/neuro/BrainSuite) or Freesurfer (google it)

2°) Is it possible to average several events when I’m using the EEG recording in order to localize the activated area in response to the evoqued potentials ?

yes: you need to epoch your EEG data about the same category of events, using an identical time window: this will import EEG data in your BST database. You can then drag and drop these trials in the process box at the bottom of the main BST window, then go to Process > Average and it will compute the event-related average of all trials/epochs.

3°) what is the best way to have a good noise covariance :

It’s always a bit tricky in EEG because there no such thing as an ‘empty-room recording’. We recommend you either use the baseline of your epochs, but for epileptic data, it’s questionable because the brain is always epileptic, right? We have recently added an option to define the noise covariance statistics as ‘identity’, ie relatively uninformed therefore with limited bias in selected data segments that best represent noise fluctuations of no interest.

3.1°) Full VS diagonal matrix

See above: I recommend you use the identity matrix option in your case

3.2°) What is the best time windows to apply the noise covariance and what is the best conditions (quiet patient, closed eyes, sleeping patient)

see above. The noise model applied to the entire data where you want to estimate sources.

4°) How can I epoch my EEG data ?

You need to define events of interest in your raw data (spikes for instance). This is possible with the BST raw editor. Once you have marked events, you can epoch the data in a time window about each marker. This is described in the tutorial Francois pointed out.

Let us know if you need more help.

17

Hello Sylvain,

I read with attention and I will work on my data with your advice.

Gael KUHN.
Student in Master of Biomedical Engineering and Drug Science.
Nancy, FRANCE.