Hi FSL experts:
One of the post from Christian said that outer product of melodic_oIC and melodic_mix will generate the signal of ICA spanned space with the time-course mean removed at every voxel.
I was trying to compare it
with the original time-course mean removed signal from the data. What I did was:
(1) calculate the global mean of 4D data from each subject by calculating the mean signal across the volume and time points within the mask.nii.gz.
(2) Divide the 4D data by the subject's global mean (i.e normalize the global mean of each subject to 1).
(3) Subtract the mean over the time points at each voxel of the subject.
I got the mean removed signal, which had similar contrast to the signal of ICA spanned space (generated from the above outer product).
I
have two questions: (1) what is the global mean value that MELODIC was default normalized to? (2) Even I multiply by a scaling factor, the residual signal (mean removed signal - ICA spanned signal) still has the brain structure, which is not noise-like signal? Could anybody tell me where my calculation for the mean removed signal is off from MELODIC procedure?
Your help is greatly appreciated!
Thank you, Weiying
|