I am performing a fairly routine analysis where I am aligning a T2 FLAIR to an MNI FLAIR template (first reference to subject, cat_matvec to get the inverse, then applying to get subject to reference). Within 3dAllineate, I am using the -cmass option to get a better alignment. After this initial alignment, I am applying the affine used to warp the subject to reference to an ROI to get it in the same space as the aligned FLAIR. The ROI is a binary mask so I am using NN interp and final to try to preserve the values. However, when I apply the transform to the ROI, it seems it is using a different interpolation than specified, as the value range changes to 0 to 0.00024 (instead of 0 to 1). This seems to be a result of using the cmass option, confirmed by using the command 3dCM and obtaining identical results. Is there a way to use the -cmass option and maintain NN interp? I do get better alignment with it, so I would like to keep it if possible, but some of the values in the mask are getting “smoothed over” with standard interpolation.
Thanks for any guidance!