Hi FSL community,
As pointed in the pioneering TFM paper, temporal ICA can be directly
performed on voxelwise fMRI data (without a prior spatial ICA
decomposition). The independent components will be the temporal
functional modes which activate/deactivate over time, and the mixing
matrix will encode the TFM spatial maps.
Now, considering that each spatial map is in the voxel domain and
encodes involvement of each voxel in the total signal of the
corresponding mode, can this map be used to obtain a metric like a
"correlation/covariance between a pair of given voxels specific to a
single mode"?
As it seems that the amount of shared variance in temporal signal of
voxels is somewhere in TFM spatial maps, I wonder if multiplying
spatial map values of any two voxels will result in a covariance
metric corresponded to the given functional mode.
Thanks in advance,
Sourena
|