About connectivity parameters for Wpli and -lagged Coh

I see that it is now possible to select different options for connectivity analysis:
Hilbert, FFT or Morlet

the question is: How do I know which one to select? (I mean, before that, default and specific parameters were automatically selected, and less confusing for me :face_with_peeking_eye:)
I mean, I want to do an analysis as close as possible to the original (published) analysis, but I don't know which was the original method for the Wpli or the Lagged Coherence.

Thanks in advance!!!

ps: this so that the results are consistent with those obtained by others and are reproducible.

Brainstorm philosophy usually tries to minimize options and keep things simple for users, but the choice of TF transform varies a lot between publications, and I find many of the "original" connectivity methods don't say much about that. The formulae are often used with the same symbols representing FFT or Hilbert transforms in different publications, which is confusing since it's not exactly mathematically equivalent, though in principle it's the same idea. So I'd have to say sorry, but there's no one choice that's "right" for every case. If you're trying to reproduce something from a specific paper, then try to determine what they did. Otherwise, it depends what time and frequency resolution you're looking for. FFT gives full frequency, no time (or a few time windows); Hilbert full time few frequency bands; Morlet in between. And there's even more flexibility with the windowing options.

Sorry again for not being able to give a very simple answer. I hope it helps.

Thanks for the answer!!
that helps a lot actually!

but I think there may be a bug:

everytime I used Hilbert TF on WPLI or Lagged Coh, there is no way to export the file to Excel spreadsheets to get the connectivity info

** Error: Line 142: export_timefreq (line 142)
** Cannot export connectivity matrices that are changing both in time and frequency.
** Call stack:
** >export_timefreq.m at 142
** >bst_call.m at 28
** >tree_callbacks.m>@(h,ev)bst_call(@export_timefreq,filenameFull) at 2038


It may be a missing feature more than a bug. Not sure if Brainstorm has a convention for exporting multi-dimensional arrays... In the meantime, you can just right click on the result in the Bst tree and select "export to Matlab". Once it's in a Matlab variable, you can manipulate it and copy/paste or export it in other ways.

Once @Raymundo.Cassani is back, maybe he'll have a more specific answer about this error.


1 Like

I see...

Thank you for the reply, really :slight_smile:

@tourette95, the exporting (to Excel) was designed to export one Excel sheet with one NxN connectivity matrix. In your case, these metrics are time-resolved (windowed) and frequency-resolved (spectral) metrics, thus, there is an NxN matrix for each window and for each frequency point. As @Marc.Lalancette mentioned, this feature is not supported in Brainstorm yet.

As such, for the moment, if you want to export the a time-resolved spectral connectivity metric, it is necessary first to extract one point in time (process: Extract > Extract time), and the use the Export to file indicating which frequency is the one to be exported.

thanks for that