Self-connectivity not 0 when running 1xN wPLI from Scouts

Hi!

I have just run a 1xN wPLI analysis on 128 channel EEG data co-registered to default brainstorm anatomy. I set 'src_rowname' to 'rostralmiddlefrontal L' on the DK atlas as my seed. I also did this with the right side.

I noticed on the 68x1 connectivity matrix, the self connectivity is not 0 on either of the seed analyses (it is 0.00x , however so are a couple of the other connections).

Is this an indication that I have run something wrongly?

Thank you!

Which Scout function options are you using?

When computing connectivity at the Scout level, there must be a dimension reduction from sources to scouts. This reduction is called in Brainstorm "Scout function" and it can happen before the computation of the connectivity metric, or after its computations. Consider these cases of computing wPLI for a Scout and itself.

If the Scout function is applied before, wPLI must be 0 for the Scout and itself, as time series in the scout sources are aggregated, and then wPLI is computed.

However, if after is selected, wPLI is computed for all the combinations of time series in the scout sources, then, the Scout wPLI is computed aggregating all the other wPLI. So it is not guarantee that wPLI will be 0.

Check this link for further documentation in the Scout-level connectivity:
https://neuroimage.usc.edu/brainstorm/Tutorials/CorticomuscularCoherence#Coherence_Scouts_x_Scouts

Hi Raymundo,

Sorry for my delayed reply.

I used the Desikan-Killiany atlas to extract 68 regions. I extracted the time series for each of these scouts and then computed connectivity between the left rostral middle frontal gyrus and the rest of the soucts using the 1xn wPLI function.

Does this mean that I have applied the scout function before and therefore the wPLI should be 0 in the wPLI matrix (which it isn't)?

1 Like

Hi again Raymundo,

I have just realized the discrepancy.

When I ran the wPLI, I first extracted values (time series matrix for each source, using absolute values), then did wPLI on the matrix. However, I realize now I can do wPLI straight on the scouts file (dSPM in my case). I see now I have the option to apply the scout function before or after.

My question is, is it preferable to run the wPLI analysis straight on the scout file, rather than extracting the values first?

Thank you,
Athena

Hi @astein
what do you mean by " extracting the values first?"

Hi @tmedani

I previously ran the extract values function first, which generated a matrix with time series' for each scout, which I then ran the wPLI function on (see pic)

I see that both approaches will lead to the same results [while using the same options at each step].
In your first method, you save the matrix to the database and then compute the wPLI,
in the second method, you can directly compute the wPLI. If you are interested only on the wPLI, you can directly use the second option.

@tmedani thanks for your answer. I'm still a little confused about my original question on this thread, though - if I first extracted the time series across 68 scouts (in in the matrix), and then ran wPLI on that matrix, why would the diagonal of the wPLI not come out as 0?

Does this response fully address your case: Self-connectivity not 0 when running 1xN wPLI from Scouts - #2 by Raymundo.Cassani
or are there additional details or issues that remain unresolved
I need to run an example and double-check your issue.