Group analysis statistics

Discussions
1

Thank you Raymundo

very helpful.

I have a few more questions

  1. Since I uploaded 62 subjects I have to create 3 different groups. How can I do this?
  2. I have to run analyses between 3 groups in term of power spectrum density and connectivity, how can I do this? How can I set statistical design?

Best regards,

Alberto

2

There isn't a way in the protocol tree to group subjects. However, as Subjects are alphabetically sorted, you could rename the subjects to reflect their group, e.g., A_Subject01, B_Subject02, and so on.

The statistics methods implemented in Brainstorm are designed to compare between two populations.
There is a detailed explanation of the available methods and options in this tutorial:
https://neuroimage.usc.edu/brainstorm/Tutorials/Statistics

PS: @albertobenelli, this post moved to a new topic, as it was an unrelated to the previous one

3

Thanks Raymundo,

I saw that if I drag the subjects of the first group, the second and the third (separately), into "Process" and click "Run" I can do the "Average" of all files creating groups. What I didn't understand is how actually this average is done in terms of source and EEG. What is the rational? What option should I do to get the average of the subjects?
Therefore, for statistical analyses I can compare two groups at a time but I didn't understand from where doing this and where to see the significant differences in power and connectivity between the compared groups, directly in the source. Could you clarify this point?

Thanks a lot for all your help,

Alberto

4

Hi Raymundo,

I saw on the tutorial web page that is mandatory having "lh.sphere.reg" and "rh.spphere.reg" for each subject in order to run the group analysis. I checked, based on tutorials, looking inside (subject1/cortex_266360V) > MRI registration > but I don't have the option "View FreeSurfer sphere". I tried to upload sphere manually using the command Load FreeSurfer sphere and the error "The number of vertices in the surface (266360) and the sphere (132868) do not match" appears. How can I fix it?

Best,

Alberto

5

The option is now called Display registration sphere/square, do you see this option?

This is correct as surfaces in Brainstorm contain Left and Right hemispheres while surface files from FreeSurfer are either Left or Right hemisphere.

EDIT: The menu names has been updated in the tutorial

6

This is described in the Statistics tutorial. There it illustrates comparisons between two groups for sensor time series, sources and TF maps. In the tutorial it is two condition groups, but the same applies for subject groups.

7

Thank you so much, I can see everything.

Alberto

8

Thank you Raymundo,

how can I find the voxel coordinates where the eeg signal come from? is there a Matlab matrix?

Alberto

9

Hi Raymundo,

How can I be sure that the source analysis went well? Is there any further step I can do to prove this?

Best,

Alberto

10

Can you clarify what is goal here?
The voxel source activation is not saved. It is interpolated from the volume grid, to the MRI volume on the fly. The activation values for the source volume grid are can be obtained with the link or source file, as described in here:
https://neuroimage.usc.edu/brainstorm/Tutorials/SourceEstimation#On_the_hard_drive

11

Hi Raymundo,

Thank you for your helpfulness.

What I'm trying to figure out are essentially three things: the first is if the source analysis went successful. Secondly, how to make groups up, with the subjects I have. Should I make an average source for the group? an average eeg file for the group? etc. Also, I would like to understand how to extract the power and connectivity values, after generating the source, from the Matlab matrices that are created; also, if this can be done by frequency bands: extract power and connectivity for delta, theta, alpha, beta and gamma separately.

Best regards,

Alberto

12

Hi @albertobenelli,

:bulb: Please post any further question that is not related to the title of the thread (Group analysis statistics) an a new post. This will help other forum users to find the information.

Source estimation methods have been validated with phantom experiments, with other modalities and by diverse experimental protocols. I recommend to you refer to the literature on the field, you can start with:

Electromagnetic brain mapping, S. Baillet et al

If you are talking on creating an "Folder" with a subset of Subjects, there is not a way to create this Brainstorm, it is not needed. However the available tools allow you to analyse the data using subsets of subjects as described above.

This is more a methodological question, it depends on what is your research question.

If you mean spectral power, check this tutorial
https://neuroimage.usc.edu/brainstorm/Tutorials/ArtifactsFilter
(It also described the use of frequency bands)

For connectivity, there is an entire tutorial on the topic:

If you haven't we strongly suggest to follow the Get started tutorials with the provided data to get familiar with the different process and options in Brainstorm. Moreover, in most of the tutorials relevant references are provided, there you can find more information.
https://neuroimage.usc.edu/brainstorm/Tutorials#Get_started