Resting-state EEG and source localization

[danielykh]

Hi, I came across some problems regarding the source localization of resting-state EEG.
My study is to obtain the source map of different microstates. So, I used Cartool for microstate analysis, and concatenate every segments of resting state EEG data labeled the same microstate for each subject, and got 10 (subjects) * 2 (conditions) * 5 (microstates) resting state EEG segments.
I used the default ICBM152 template, and have already done the source localization with sLoreta and averaged the source map in the time dimension for every EEG segments. What I want to do next is to average the source map across subjects for each microstate. How can I achieve it?

In the publication, it described 'For each microstate, subject-average source maps were z-scored, rectified, and group-averaged'. I tried to conduct the z-score standardization with all file set as baseline, it showed invalid scale in viewer.

Best regards.

[Raymundo.Cassani]

The error seems to be in the incorrect use of normalization:

@danielykh

I used the default ICBM152 template, and have already done the source localization with sLoreta and averaged the source map in the time dimension for every EEG segments.

With this step, each of your files is now a time sample (the average).

@danielykh

I tried to conduct the z-score standardization with all file set as baseline, it showed invalid scale in viewer.

Normalization is applied to one sample, thus its std = 0, then the z-score is undefined, thus the invalid scale.

The idea behind normalization with z-score is standardize across subject as they might have measurements on different magnitudes. As such, for each subject, you may want to normalize these microstate averages with respect to another segment, e.g., a resting state segment.

[danielykh]

@Raymundo.Cassani

The idea behind normalization with z-score is standardize across subject as they might have measurements on different magnitudes. As such, for each subject, you may want to normalize these microstate averages with respect to another segment, e.g., a resting state segment.

Thank you for your detailed explanation! Should I follow these steps:

1. For each subject, calculate the time-averaged source map for each microstate source map.
2. Normalize the averaged microstate source maps by using the subject's resting-state segments as the baseline (i.e., compute the time-averaged source map from the raw EEG data (before microstate segmentation) and use this as the baseline).
3. Perform group-level averaging.

Is this workflow correct?

Furthermore，when performing group-level averaging, should I perform first 'Pre-process > Absolute value' for each source maps and then use 'Average>average files(z-scored time-averaged individual microstate source maps files)' for each microstate? Do I need other operations? Since I used the ICBM152 template, it may not be necessary to project the source map on the template..?

Additionally, the publication describes ' Group-averaged maps were thresholded to activations above 95th percentile, as suggested in fMRI conjunction analysis ', is this operation performed using Test > Threshold by percentile? (as shown below)

[Raymundo.Cassani]

@danielykh

1. For each subject, calculate the time-averaged source map for each microstate source map.

Sounds good

@danielykh

2. Normalize the averaged microstate source maps by using the subject's resting-state segments as the baseline (i.e., compute the time-averaged source map from the raw EEG data (before microstate segmentation) and use this as the baseline).

The definition of the baseline depends on your hypothesis. I just mentioned "a resting-state" as example. As this is a common choice so the z-score indicates the changes with respect to it.

@danielykh

3. Perform group-level averaging.

This is our general recommendations for sources group analysis:
https://neuroimage.usc.edu/brainstorm/Tutorials/Workflows#Average:_Group_analysis

@danielykh

Furthermore，when performing group-level averaging, should I perform first 'Pre-process > Absolute value' for each source maps and then use 'Average>average files(z-scored time-averaged individual microstate source maps files)' for each microstate? Do I need other operations? Since I used the ICBM152 template, it may not be necessary to project the source map on the template..?

This is correct, as you are using the Default anatomy for all subjects, projection is not needed.

@danielykh

Additionally, the publication describes ' Group-averaged maps were thresholded to activations above 95th percentile, as suggested in fMRI conjunction analysis ', is this operation performed using Test > Threshold by percentile? (as shown below)

Yes, that does the job. It sets to 0 all the sources below the indicated percentile.

• If your sources are already rectified (abs values), the Sort the absolute values... does not have any effect.
• Check the Signals checkbox
• If your sources are only one sample (there is not time), the status of the Time checkbox is irrelevant. Same goes for the Frequency checkbox.

[danielykh]

Thanks for your detailed explanation! It really helps a lot!

[danielykh]

Hello, I still have some questions. For group-level averaging, what is the most suitable method for standardizing the source maps of resting-state EEG? (Since z-score requires selecting a baseline segment, I believe there is no appropriate baseline available.)