Power spectrum on ICA components

[monaghaheri]

Hi dear all

I am interested to remove ICA component by some cues other than time series and scalp topography. I have noticed that there are some other cues like power spectrum, percentage of each component variance ,etc in EEGlab.
I have read in one of discussion in that time( 2016) there wasn't power spectrum feature for each component.

ICA - spectrum

Hi, Is there a way to visualize the spectrum for the ICA components already computed? Thank you, Have a great day. Véronique M.

If I can access to these additional features in Brainstorm, would you please guide me?

[Francois]

Unfortunately, exploring IC in spectral domain is still not possible in Brainstorm.
One option is to do the pre-processing of your signals in EEGLAB (to explore and remove some components identified in frequency representations) and then epoch the resulting .set file in Brainstorm.

[monaghaheri]

Hi Francois,
Thank you for your suggestion.
I have another question, not related to this topic but around ICA components:
How can I demonstrate event trigger by line on component's time series?

[Francois]

I understand that you would like to see the event markers on the IC time series instead of only the EEG signals. The limitation being that if there are multiple figures on the same recordings, only the first figure opened will show the events.

First, update Brainstorm to get a small fix to make the procedure easier:
https://github.com/brainstorm-tools/brainstorm3/commit/4fe430cd3cd2d526ffd025059348d44504ec0001

Then, you should be able to achieve this with this procedure:

1. Open the EEG recordings => You should see the event markers, as lines or dots depending on your personal configuration
2. In the Record tab, menu Artifacts > Load projectors as montages
3. In the montage selection menu (in the Record tab, or in the display menu of the figure), select the ICA decomposition you want. You may have to modify a lot the vertical scale (right-click + mouse up/down)

[Marc.Lalancette]

+1 for this feature. I also wanted to look at the spectrum of components (in my case SSP).

It is actually doable, but requires a few steps:

1. Load projectors as montages.
2. You then have to edit montages, save this "tmp" montage and reload it. You can rename and edit it now if you wish, say if you only want to keep a subset of components.
3. Import the data (without epoching or applying SSP).
4. Run the process: Standardize, apply montage.
5. You can then finally compute the power spectrum.

This could be simplified by allowing raw files as valid inputs for the process "apply montage". We could then skip steps 2 and 3. It might also be useful if saved projector files were compatible with montages, i.e. if they could directly be loaded as montages.

Finally, note that "Load projectors as montages" should probably be greyed out for imported data files, since it doesn't seem to do anything.

[Francois]

This could be simplified by allowing raw files as valid inputs for the process "apply montage"

Far from being trivial.
Code suggestions are welcome.

It might also be useful if saved projector files were compatible with montages, i.e. if they could directly be loaded as montages.

Sure. This will not be a priority in my list of developments, but you're welcome to work on it.

note that "Load projectors as montages" should probably be greyed out for imported data files, since it doesn't seem to do anything.

I'm not sure I understand this comment. If the recordings were imported with SSP or ICA projectors applied, you can visualize the ICA/SSP components for these files (even after averaging).

[Marc.Lalancette]

Far from being trivial.

Ah, I would have thought it would be rather simple since you must already create a new channel file even for imported data. Maybe I'll have a look.

If the recordings were imported with SSP or ICA projectors applied, you can visualize the ICA/SSP components for these files (even after averaging).

Ah, I think I was looking at a file without SSP applied. In this case, "Select active projectors" is greyed out, but not "Load projectors as montages".

[Francois]

@Marc.Lalancette

I think I was looking at a file without SSP applied. In this case, "Select active projectors" is greyed out, but not "Load projectors as montages".

At the creation of the menu, we don't know whether there are projectors available or not. The availability of the menus is only defined by the distinction continuous/imported.

I will not change the availability of the menus, but I added a feedback message when clicking on this message, to tell you how many montages were created.

GUI: Load projectors as montages: Display feeback message

[Marc.Lalancette]

Thanks

At the creation of the menu, we don't know whether there are projectors available or not. The availability of the menus is only defined by the distinction continuous/imported.

If the recordings were imported with SSP or ICA projectors applied, you can visualize the ICA/SSP components for these files

I thought that meant that "select active projectors" was available in that case, but I was wrong. How did you mean we can visualize these components on imported data? I was able by loading projectors as montages on the raw file, keeping it open, then opening the imported file. (But if the raw file is closed first, those temporary montages are no longer available.) Is that the way you had in mind?

[Francois]

Consider the introduction tutorials, the deviant average in run01.
It is imported+averaged from a raw file on which we had removed blink+cardiac SSP components.

The menu "Select active projectors" is disabled, because we don't offer interactively the selection of SSP projectors on imported files.
The menu "Load active projectors as montage" is however available (without checking if there is actually anything to be used in the ChannelMat.Projector field).
If we click on this menu while the deviant average is loaded, we get access to the SSP decompositions as new montages. The first component is flat because it has been suppressed from the signal at the import time.