I am currently trying to extract scout time series in the pars opercularis from an unconstrained LCMV beamformer source estimator. When I extract these time series I have the option to get the norm of the three orientations (producing only positive values) or obtain three separate time series, one for each orientation.
In the tutorials it is mentioned that dipole scanning can be used to select the optimal dipole orientation. How can I do this to obtain a single scout time series (not the norm) representing the optimal dipole orientation?
I tried applying dipole scanning and simply extracted scout time series, but I still obtain the norm version of the dipoles.
Thanks Francois for replying so quickly. I will await their responses. I wanted to ask a side question. There are some bad channels in my epoched data which I have removed. Do I (should I) reinterpolate them manually or would the source estimation algorithm do it automatically? Also, is there a favourable referencing method (e.g. average reference) that I should invoke?
No need to interpolate between bad channels (it does not add more information to source modeling) and no preferred option for referencing (it is simply a matter of visualization; Brainstorm performs source modeling with whatever EEG reference is specified).
Concerning your initial question, @jcmosher is the most qualified to answer your question. What I would recommend is that you extract the 3 LCMV time series from a given location and perform their PCA decomposition. I believe the first 3x1left vector would correspond to the best dipole orientation. (you need to do this using your own Matlab code)
Thanks for the advice Sylvain, I will certainly try that. If I want to look at scout signals on a trial by trial basis, would you recommend that I use Minimum Norm Imaging > current density map > constrained > median eigen value?