Hi,
I'm doing some frequency components analysis with deep brain mixed head model. Using these 3 measures of source estimation options, I'm getting different magnitude for the FFTs in cortical and subcortical area. Specifically, with the current density map, magnitude of the FFTs in cortical areas is larger than subcortical areas, for dSPM and sLORETA, it's the opposite. From the tutorial, I understand these 3 measure have different bias toward deep structure, but I want to make sure the degree of differences I see among these 3 measures still make sense.
For example, using current density map, cortical FFT magnitude is about 10 times larger than the subcortical FFT magnitude.
using dSPM, cortical FFT magnitude is about half of the subcortical FFT magnitude.
using sLORETA, FFT magnitude are similar between cortical and subcortical areas.
Another question is: does number of vertices of each ROI affect this ROI's FFT magnitude? When I look at the FFT magnitude of each ROI, should I take into consideration of the size of each ROI?
Thanks,
Yi
Using these mixed head models, different types of source models are used for different brain regions. The cortex typically have one dipole at each cortical location (constrained dipole orientation), most subcortical structures use volume grids with 3 dipoles at each location (unconstrained dipole orientation). The density of dipoles is therefore very different between different brain regions, causing important variations of estimated current density at each dipole.
Together with variations of depth, there are too many parameters varying to study the difference of FFT power between cortical and subcortical regions.
In general, I would advise not to try comparing directly the activity (current density, dSPM or power for a given frequency) between different brain locations. What is interesting is to detect the significant differences between two conditions (comparing one brain location across conditions).
Note that statistics are also complicated to compute for unconstrained or mixed models. In many cases, our recommended approach is to stick with source models with constrained dipoles orientations.
Another question is: does number of vertices of each ROI affect this ROI's FFT magnitude? When I look at the FFT magnitude of each ROI, should I take into consideration of the size of each ROI?
The more you average things that are unrelated, the lower the results.
Try using regions of similar size (ie. similar number of dipoles and similar spatial extension). Pay attention that comparing amplitude for constrained sources and unconstrained sources is not very relevant.