What I did is to localise (constrained, dSPM) the time course of beta coefficients characterising the linear relation between the brain signal and a variable of interest.
You can't apply the Brainstorm inverse routines to anything but MEG or EEG signals. Multiplying regression coefficients with the inverse models computed with Brainstorm will not give any meaningful results.
See: Can I do source estimation for a single spatial map (not time series) - #2 by Francois
I'm not sure I understand this sentence correctly. If this comment is off-topic, please ignore it.
How could I obtained the correct signal direction, in other words, how could I tell where positive and negative betas are encoded in the brain?