Wpli and connectivity metrics validity/publishability

Hello Francois, hoping all is well. Thanks as always for an excellent and useful software.
I've sent this as a private note as it is unclear if useful to send in the Forum.

In regards to wpli in BS below: [unclear how this testing/checking relates to main current BS development goals, but surely having clarity/validity regarding the connectivity metrics is important BS to continue to grow as a go to for E/MEG analytics.] If there is no timeline and if there remain many unknowns, that is fine of course, but important to know.

My apologies if my questions are naive, and I presume they reflect targets for BS developers anyway.

  1. it seems that changes were made recently on github to BS code for wpli - and it's not clear what the resolution was there
  2. however, from notes/exchanges here in the Forum, and in the Github notes
    it remains unclear
  3. whether wpli estimates were valid before last changes on Date X
  4. whether new wpli estimates are correct/valid after Date Y
  5. what's the difference between the two versions
  6. how are these BS versions different specifically from the published wpli, dwpli,
  7. were the tutorial pages updated to reflect or document these changes?

The above also brings up worries about any other connectivity metrics in BS
We would love to lean on and trust BS for large sample data analyses,
but without more (and consistent) clarity on metrics details such as the above, it suggests that it's up to the users of the software to do validity checks on any/all connectivity measures being used,
to dig into the unique code for each, compare with code/methods in other software,
and then make best choice possible.

One last question for you: Do you think the field will have to do something like the below for each metric of interest across softwares ? I am sure you are familiar with other similar recent work, e.g., MNE across softwares.
See figure 5 for example here:

The forum is the perfect place for this discussion.
These are questions for signal processing specialists:
@danielemarinazzo @hossein27en @leahy @Sylvain @Marc.Lalancette @Raymundo.Cassani


The above also brings up worries about any other connectivity metrics in BS

Please be more specific.
Which of the metrics documented in the tutorials below do you think are not documented properly?

Reference for wPLI discussion: Weighted phase lag index

Hello thank you Francois for your note!

I am not expert on spectra and connectivity, and may not be able to successfully compare BS code and published algorithms. will be happy to review documentation closely and try to respond to your extended query as well as best as possible.

what drew me to ask question was wpli discussions and changes in forum and github. For the time being was mainly curious if wpli is OK in BS Feb 20 2022, and whether we can trust wpli from october 2022, how they may be different if developer changes were done since 2022, and if further changes expected.

We would love to trust BS 100% but we understand due diligence and verification is rightly in large part up to BS users.


Indeed, wPLI and associated phase-based metrics still need to be fixed in BS. As discussed on GitHub:
Update bst_connectivity.m by danielemarinazzo · Pull Request #583 · brainstorm-tools/brainstorm3 · GitHub and
This is on my to-do list. Work had started, but then got delayed. I apologize for the resulting current state of things, including the lack of clarity on the situation.

For BS team: This is the next item on my BS list after PCA. I hope to get back to it soon. I could open a PR with what I had done so far if that would help, e.g. if someone wants to work on it right now.