Dear Afni experts,

We have a study that consists of 4 functional runs and we are faced with the issue of registration of functional data across runs where there is a possibility of large shifts in head position across the different runs (1hour in scanner, 8mm shifts common).

The quick solution I assume would be to use something like align_epi_anat.py to align all runs to a single reference time point. My concern is that because the study is looking at practice effects over runs, differences in the magnitude of alignment would translate into differences in the spatial smoothing induced by the alignment procedure. For example, if Run2 is around 1mm translation from the reference point in Run1, and Run4 is around 5mm away from that reference point, the alignment of Run4 might introduce more spatial smoothing. This will result in different (inadvertent) smoothing for the different runs, affect tSNR and could result in different results for these runs.

The only potential solution I can think of, and I don’t have implementation details, is to align the anatomical SPGR to each functional run separately, and then project the functional data to a SUMA surface and mesh matching that run (i.e., the freesurfer-generated surface will be aligned to each of the 4 runs separately and data from each run will be project to a "common mesh" where vertex numbers correspond in anatomical space).This sounds a bit convoluted. I don’t know if it’s possible to do something like this staying in volume space, which would be easier. i.e., align the anatomical SPGR to each run separately, then resample the SPGR grid to the EPI grid (3dfractionize) and then, using some hack yet-unclear to me, that will allow inheriting a common indexing scheme from the SPGR to the EPI dataset so that anatomical location X in Run1 would correspond to the same anatomical location in Runs 2,3,4.

If anyone has thought of this issue and has suggestions, much appreciated.