 |
 |
Hi, It all depends on what you mean by "registered". If you change the FOV, or crop the output, then you might have the rotations applied but it isn't clear how you would relate a voxel between your anatomical image and your resampled functional image. Do you want to keep the NIFTI mm coordinates the same in both? Or the centre of volume? By default we keep the corner of the FOV of the same, so that an FSL identity transform relates the two images together. One thing you might find better is to apply the offset after the rotation. Currently, using the -init option, you apply the offset first, whereas you might find that things work better to apply it afterwards, by using the concat option in convert_xfm, with the offset as the <mat_BtoC> argument (or you can just add values directly to the fourth column of the matrix). Applying the offset after the rotations is more likely to give you consistent results and will hopefully do what you want. All the best, Mark > On 3 Aug 2017, at 18:33, RC Lapate <[log in to unmask]> wrote: > > Dear all: > > I am currently trying to register functional EPIs to the subject's anatomical scan while keeping the voxel resolution native. The resulting registered image is cutoff, because the FOVs are different and the COVs are not matched (cf https://fsl.fmrib.ox.ac.uk/fsl/fslwiki/FLIRT/FAQ#Can_I_register_to_an_image_but_use_higher.2BAC8-lower_resolution_.28voxel_size.29.3F). > > I am familiar with the solution of adding a translation to the right hand column of the transformation matrix (where the offset to be added should be half of the difference in the respective FOVs). I have indeed tried that (e.g. flirt -in pe1 -out p1_2_anat -ref example_func -applyxfm -init example_func2highres_offset.mat, where example_func2highres_offset.mat is the matrix with the offset added), however, that solution is not working for the dataset I am presently working with (where EPIs are isotropic 2.5mm FOV: 84x84x52; the anatomical is isotropic 1mm FOV 160x256x256), and the resulting registered image is still cutoff. > > As a possible alternative approach, I have also created a reference image (using ApplyXFM) that is 2.5mm isotropic (like the functional) with the anatomical's FOV (160x256x256), and used that as "ref" (instead of example_func). Now, while the resulting registered image is not cutoff, it is of course minuscule, and located in the leftmost corner (i.e., necessitating cropping for most visualization etc purposes)... > > --Is there a better solution? > > Thanks for your help, > > Regina