Hi Rick,
How can I preserve volume in the result? This would definitely solve my problem.
I am extracting raw time courses (averaged within ROIs) for specific time ranges (varying from subject to subject) denoted by 1D files, using 3dROIstats where the mask has ROIs denoted by 1 through n. Previously, I had a +tlrc space mask where each ROI had the same volume ( 12 voxels, 12mm*12mm*12mm ) and I used 3dfractionize to transform this mask into each subject's orig space, where each voxel is 4mm*3.4mm*3.4mm. This was entirely sufficient to extract a mean original time series in each region, but the number of voxels in each region ranged from 25-~35 -- with a clip level of 0.35.
This was sufficient for a while, but now I want exactly 27 voxels in each ROI (i.e. each subject's mean time series includes the same number of voxels centered at the same anatomical point) so I made a +tlrc mask with 1 point per ROI, and then used 3dcalc with differential subscripts to expand each ROI by 1 voxel in each direction on the +orig transformed mask. For obvious reasons, 1 point per ROI in the +tlrc mask did not necessarily translate into 1 point per ROI in the +orig mask (as I was hoping would happen), so I still don't have exactly 27 voxels in each ROI.
In general, it is clearly not possible to preserve the number of voxels; but in this case, each ROI is spaced sufficiently far apart that I thought it would make sense to aim for one 4*3.4*3.4 mm^3 voxel in the +orig mask per 1*1*1 mm^3 in the +tlrc mask. Thanks for your help -- volume preservation sounds like a fantastic solution, and I need guidance to implement this.
Thank You!
CS