Creating protocol, design


When creating Protocol, is it OK to include subjects separately like in the plot bellow? Channel numbers are the same across subjects.
In this case I have to calculate head model for each subject!?


When I am averaging conditions from 2 subjects, new file is created Group analysis, intra-subject (plot above).


1 Like

Hi Niko,

Can you clarify what you're trying to do here? This will help us answer you better.

Brainstorm supports having multiple conditions and runs per subject, so splitting your first subject as Subject01 and Subject01_2 is probably not required. Head models can be copied from one folder or subject to the next if they have the same anatomy. Right click on the head model -> File -> Copy, and then right click on the destination folder -> File -> Paste, or use CTRL + C and CTRL + V.


Thanks Martin, Please ignore Subject01_2. I am creating separate file per subject, and each subject has 5 conditions. Anatomy is the same across subjects. Should I create separate Subject files, and copy the head model like you recommended or just create 1 Subject file and include all subjects and conditions there?

Just as a side note, when I am uploading preprocessed .set file (continuous EEG recording), I can not name the file, it appears as a Raw.


Thank you Niko, that's clearer.

Since the different conditions within the same subject have the same channel and anatomy information, it's fine to share the channel and head model files, if you are not using ICA or SSP cleaning. Right click on your subject -> Edit subject -> Default channel file: Yes, use one channel file per subject. The channel and head model files will appear under "Common files" rather than a specific condition. It's also fine to duplicate the head model if disk space is not an issue.

You cannot rename raw files in Brainstorm ("Link to raw file"), but the files called "Raw" in your screenshot are actually imported recordings. You should be able to rename them if you right click -> File -> Rename (or the F2 keyboard shortcut).

I hope this helps!

Thanks Martin.

I wonder which of these 3 versions looks better?

1 - as shown in figure bellow, to switch to global channel file for whole protocol. Note that channel info, also number of channels are the same across subjects. I am not doing any preprocessing, just uploading analyzed .set files.

2 - as shown in figure bellow, to create just 1 subject file and to include all subjects here.

3 - similar to 1, but instead of global channel file, to select one run per subject.


This is all valid, chose what is more convenient for you.
Never share channel files if you are using SSP or ICA cleaning for your data.

1 Like

Thanks Francois.

I would like to ask if I am using other software for preprocessing (including ica), can I choose to share the location file when I import the processed epoch file into brainstorm?
I will not do ica in brainstorm later.

This depends on the file formats you are using for the preprocessing.
If you import preprocessed files from EEGLAB, the sensor positions should be carried out correctly.

Be aware that preprocessing with ICA your recordings before importing them in Brainstorm may lead to less accurate source estimation. The ICA mixing matrices are used in Brainstorm in the inverse computation, but can't be imported from EEGLAB.

I am using mne for data analysis and exporting to a fif file.

Do you mean that the ica data will have an impact on the brainstorm traceability?

So is it necessary to do ica analysis in brainstorm?

Do you mean that the ica data will have an impact on the brainstorm traceability?

If the ICA is done in MNE, I'm not sure the ICA mixing matrices are saved to the FIF file and loaded by Brainstorm. You can check this by looking into the Projector field of the channel file.

So is it necessary to do ica analysis in brainstorm?

The ideal would be to have the ICA mixing matrix applied to the head model before inversion, otherwise the forward model that is used for source estimation does not match the EEG data on which the inverse model is applied to.
But we never studied in details how much different the results you obtain are.
This is something you can test with an example dataset, like this one:

@Alexandre Any advice to share?

if you whiten the data and the gain matrix with a noise covariance that has been computed from data processed with the same ICA model you are fine.