Dear AFNI Gurus,
We are using the retrots.m and the ricor procedure to remove physiological noise. We call the image correction procedure using the afni_proc.py, which creates the actual command lines (proc.$$ tcsh script).
Judging by what we see in the proc.$$ script and the fact that the entire workflow runs, we think that the workflow is working as expected.
But, the resulting .ricor files for our subjects (e.g., pb01.$$.r01.ricor+orig) -- which to our understanding are the final output of the process -- seem to have an unexpectedly large spatial smoothing factor as estimated by 3dFWHMx. We think this is a problem, perhaps induced by how the .ricor file is generated from the .errts && .polort files that are intermediate products of the retroicor process.
Here are the details below.
1. Our original time series (detrended with polynomial 3 for comparison purposes) has an estimated smoothing kernel of 4.7, 5.4, 4.3
2. The .errts file generated from retroicor has a smoothing kernel of 4.2, 5.0, 3.5 (seems to be behaving properly).
3. The .polort file generated from retroicor has a smoothing kernel of 20.9, 25.6, 21.7mm (this polort file is generated automatically from a 3rd order polynomial command line)
4. The .ricor file (which is automatically generated by adding the .errts and .polort files) has the following: 20.8, 25.5, 21.7
In sum, we start off with a smoothing of around 5mm, and end up with a smoothing of around 20. This occurs for multiple subject files. While we would expect that correct physio noise removal could result in a slight increase in spatial similarity, these value seem to be out of bounds.
Any advice, MUCH appreciated.
Best,
Oori