I'm trying to do something similar and have run into some problems. I have generated a template from high resolution data and am trying to register the data to that. I have edited the tbss_2_reg and tbss_3_preproc files to point to that template, but things still aren't working.
I have two questions:
1) In the tbss_2_reg script, there is an fsl_reg call for an FA-optimized registration that clearly uses fnirt. I assume the -FA call points to a cnf file, but does that cnf file depend on the size of the image, or is it just based on image contrast? If it does depend on image dimensions, how do I switch fsl_reg to run in flirt mode? Do I just want to just put in a regular flirt command?
2) I'm confused about the potential need to reassign voxel dimensions. My voxels are anisotropic (1.1 mm slice thickness). I assume I just need to change the in-slice dimensions to 1x1 in fsledithd? Do I need to also rescale the image matrix?