ICA automatically ignore bad segments and channels?

1

Hi,

I want to to know that if ICA or any other process in brainstorm ignore the bad channels and segments or I should again load my data without those bad information?

Thanks,
Talesh

2

Bad channels are always removed in ICA and SSP analysis, and they should.

For bad segments, it depends on the type of inputs:

  • If you compute the ICA components on epochs (imported recordings or continuous files around events of interest): the epochs including at least one event with the tag “bad” are excluded
  • If you run the ICA analysis on continuous recordings, the bad segments are not excluded because we want to keep the signal continuity.
3

Thanks for your answer

In fact, I have a continuous data with 40 minutes in which 3 segments each with during about 30-40 seconds exist. I don’t want to epoch my data but I am interested to process the data using ICA so that ICA automatically ignore those bad segments and after applying ICA I obtain a dataset without those bad segments.

Is it possible in brainstorm automatically or I should change the ICA codes (changing two script files: “ica.m” and “ica2.m”) or I should remove those segments in Matlab and then load in brainstorm (an inconvenient method for me because I should do it for 90 subjects).

Regards,
Talesh

4

There is no way to remove bad segments directly from continuous files. The main problem with such solution is that they create sharp discontinuities in the signals, and it changes the timing of the file (Brainstorm does not support any jump in time for continuous files). To avoid any confusion, we decided not to allow such solutions.

You can go around this limitation by importing your segments of interest (what you want to keep) in your database, then running the ICA decomposition on them (Process2, FilesA = imported good segments, FilesB = original continuous files).
Keeping the bad segments should not cause any issue.

If you really want to remove the bad segments: in Process1, select all the imported good segments, run process Concatenate time, then right-click on the output files > Review as raw.

5

My signal amplitude is around 2000-20000 but around the bad segments the amplitude goes around 50000-500000. I want to know your idea about applying ICA to this signal. Should I remove bad segments before applying ICA or it can not pose a bad effect on ICA performance?

6

This is something you could ask on the EEGLAB list.
Brainstorm uses directly the runica function from EEGLAB, but I don’t have much experience with it.

7

OK dear Francois, thanks a lot for your quick answering