I've used a mixed model for cortical + subcortical head modeling.
And I need a single gain matrix value per vertex, and it is possible to constrain one direction for the cortex using 'bst_gain_orient'.
However, as far as I know, unconstrained orientation should be used for the subcortex except for the basal ganglia and there are no methods for constraining one direction.
Then how can I convert the three orientation gain matrices (x, y, z) into one orientation?
Is it not possible to convert them (what about averaging or summing the three orientation gain matrices)?
If it is impossible, can I use the constrained orientation method for the subcortex the same way as I would for cortical regions?
You are correct that in Brainstorm, cortical sources are typically constrained to the normal direction using bst_gain_orient
If you need a single gain matrix value per vertex for the subcortical sources, you generally cannot directly constrain them in the same way as cortical sources. However, there are some potential approaches you can consider.
The simplest way is to consider them similar to the cortical source and apply the same transformation.
The other option is as you mentioned, averaging or summing.
Another alternative is the PCA of the three values [ perform a Principal Component Analysis (PCA) on the gain matrix at each vertex and select the dominant direction.]
Among the four options you mentioned, could you advise which one would be the most suitable for achieving accurate source localization in the subcortex?
Or do these options have comparable accuracy in source localization?
I would be grateful for your guidance on this matter.
As the orientation of the neurons in these areas is not well-defined and organized, it is difficult to determine their orientation. I suggest using PCA to analyze the data and observe the results. You can also compare the outcomes of these different approaches and evaluate them against your hypotheses.
Following your suggestion, I referred to the Braintorm PCA tutorial (https://neuroimage.usc.edu/brainstorm/Tutorials/PCA) and ran the PCA using the "across all epochs/files" option.
However, I am having difficulty finding the process to flatten the gain matrix using the PCA outputs.
I have attahced the output from running PCA for your reference.
