I have two questions about the AEC process in Brainstorm.
In Brookes et al 2012, they describe two similar processes, Average Envelope Correlation and Correlation of Averaged Envelopes. The primary differences is averaging Pearson correlation coefficients of time segments vs. averaging Hilbert envelopes of time segments and then running the correlation. Which form is implemented in Brainstorm?
In a recent paper comparing various functional connectivity metrics using MEG, Colclough et al 2016 recommend using AEC over other metrics. However, they precede this calculation with a “symmetric orthogonalisation” procedure to remove shared signal at zero-lag between network points (I assume vertices or scouts in this case).
Is this process implemented in Brainstorm, or is there some equivalent process for removing zero-lag signals to help prevent issues of spatial leakage?
Question 1
I’ve watched the Brainstorm code. To my understanding Brainstorm first calculates power envelope of the time series and then calculate correlation coefficients across these envelopes (see lines 423-480 of the brainstorm function bst_connectivity.m, that you can find within the brainstorm folder with this path brainstorm3/toolbox/connectivity). So, if you have a long continuous file and you want to obtain the Average Envelope Correlation you may want to split your recordings in segments (i.e. epochs), then calculating the AEC, ticking the option “save the Average Connectivity Matrix”.
If you want to Averaging Hilbert envelopes you should play a bit more and possibly you should do something with code (but you could easily calculate the envelope, then averaging that, and then performing the correlations).
Question 2
Concerning you second question in the Amplitude Envelope Correlation process there is an option to “orthogonalize” signals ( for details see line 447 of bst_connectivity.m).
Hope this helps, but better to wait for a response from the Brainstorm developers