Frequency bands for the Coherence NxN

neuro151 · May 11, 2021, 6:53am

We can calculate the value of different frequency bands separately when calculating PLV or PTE. When calculating the coherence, is there a similar frequency division calculation option? How to compute the coherence in different frequency bands.

tourette95 · May 11, 2021, 7:53pm

Easy: once the NxN Coherence is done, drag the file to the process tab, press Run --> Frecuency --> Group in time or frequency bands

neuro151 · May 12, 2021, 7:07am

Thank you for your reply. I think that grouping in frequency bands should be done first and then Coherence. Whether grouping in frequency bands after Coherence is the same as Coherence after grouping?

Francois · May 13, 2021, 2:12pm

PLV and PTE are methods that start by filtering the broadband signals in a subset of frequency bands, and then it computes a connectivity measures for these filtered signals.

Coherence is based on Matlab's the FFT function, you obtain your results for the full spectrum of your signal, between 0 and Fs/2.

Methods are documented here:
https://neuroimage.usc.edu/brainstorm/Tutorials/Connectivity

I think that grouping in frequency bands should be done first and then Coherence.

This does not make much sense. What you should have in input of the coherence computation is a broadband signal.

neuro151 · May 28, 2021, 2:38am

Thank you for your reply. Another question is that there are zero-values in the matrix. How to get a matrix without filtering p value?

Francois · May 28, 2021, 2:27pm

There is not much to expect from the values that are filtered out from the p-value. It means mostly that you don't have enough data to estimate the coherence properly.

If you want to change this, you would need to edit the files process_cohere1n.m or process_cohere2.m and change the option OPTIONS.pThresh = 0.05; to 0.

neuro151 · May 30, 2021, 4:59am

Hi Francois,
I edited the files process_cohere1n.m and process_cohere2.m and changed the option OPTIONS.pThresh = 0.05; to OPTIONS.pThresh = 0; Then, the matrix of results became the 1-0 matrix. Why?

Francois · May 31, 2021, 12:07pm

Then, the matrix of results became the 1-0 matrix

What does this mean?

The effect of pThresh on the connectivity matrix is visible in bst_connectivity: the connectivity value for all pValues below pThresh is set to 0.

github.com

brainstorm-tools/brainstorm3/blob/master/toolbox/connectivity/bst_connectivity.m#L300


      
                      OutHist = bst_history('add', OutHist, 'src', [' - ' OrigFilesA{iFile}]);

                  else

                      OutHist = bst_history('add', OutHist, 'src', [' -A: ' OrigFilesA{iFile} ' -B: ' OrigFilesB{iFile}]);

                  end

              end

          end

              

          

          %% ===== CONCATENATE INPUTS / REMOVE AVERAGE =====

          sAverageA = [];

          sAverageB = [];

          nTrials = 1; % for Granger

          % How to read the files

          switch (OPTIONS.OutputMode)

              case 'concat',  isConcat = 1;

              case 'avgcoh',  isConcat = 2;

              otherwise,      isConcat = 0;

          end

          

          % === LOAD CALLS ONLY ===

          % % Prepare load calls, data will be loaded in bst_cohn_2021

neuro151 · May 31, 2021, 1:23pm

Compared with the condition of OPTIONS.pThresh = 0.05, the matrix of results (OPTIONS.pThresh = 0) became a dual-value matrix (0/1). According to R(pValues > OPTIONS.pThresh) = 0; ,maybe I should change the option OPTIONS.pThresh = 0.05; to 1 ?

Francois · May 31, 2021, 2:04pm

maybe I should change the option OPTIONS.pThresh = 0.05; to 1 ?

Yes, you are absolutely right, sorry I sent you on the wrong tracks.

neuro151 · June 1, 2021, 1:54am

It's ok, thank you.