Using bst_cohn_2021.m for fMRI coherence calculation

Discussions
,
1

Hi everyone,

MATLAB's mscohere function is extremely slow when computing coherence, but I noticed that bst_cohn_2021.m in Brainstorm is much faster, so I wanted to use this function to compute coherence for fMRI signals.

My question is: is this even possible? Is it methodologically accepted? And if so, how should the parameters be set, for example WinLen?

Thanks for your help.

2

Hello @Hamidreza,

Yes, it is possible. You can certainly use bst_cohn_2021() to compute magnitude-squared coherence (MSC) using the FFT approach. The function is quite versatile and can compute coherence for the cases 1xN, NxN, and AxB (See our connectivity tutorial for reference).

Yes, the same results are obtained as with mscohere when using the same configuration.

WinLen is the length of the window that will be used to compute the PSD and cross-spectrum estimators (Sxx, Syy and Sxy). That parameter depends on the minimum frequency step you want to reach, and the available quantity of data.

Please note that in bst_cohn_2021() the estimators Sxx, Syy and Sxy are computed from the total number of windows, i.e., windows-in-file times nfiles