We would like feedback on our attempts to convert the MR time series at each voxel at each TR to a % signal change value.

Our goal is to then input the new file containing % signal change into 3dDeconvolve so that its output is expressed in % signal change.

We also want to compare % signal change at every TR at every voxel in run1 and run2.

Thus,
for a given voxel we used 3dcalc:
( x - Mean / Mean ) * 100
where x is the raw MR unit (after doing 3dTshift, 3dvolreg, 3dDespike)
This should give us percent signal change at every TR for every voxel.

Doing this gave us extremely high values that did not make sense.
Then we ran 3dDetrend on this percent signal change and removed the mean and linear drift. This resulted in values that made sense in the brain but we want your feedback on whether or not this is an appropriate series of steps.

We also tried 3dDetrend on the raw MR time series data and then converted to percent signal change. But, this method gave us extreme and odd values.


Thanks in advance.

philippe