I’m no stranger to troubleshooting alignment woes, but this is a particularly frustrating case, and I welcome any tips.
I have a dataset with about 15 T1w structural images and ~80 resting state EPIs (same brain). The T1w image is more or less AC-PC, whereas the EPIs are extremely oblique. Long story short, the align_epi_anat.py segment of my afni_proc.py pipeline is not having a great time. I have typically managed this across a dataset by figuring out the general shifts and rotations necessary to get things reasonably close to alignment with the T1w structural, and then apply 3dZeropad and 3drotate across all of the EPIs. The trouble is that the EPIs vary pretty widely in how far off they are, and in what ways. Thus, if I figure out a ballpark set of parameters to use for one scan, it will work on a dozen or so and fail on all the rest. (For the record, I did not acquire these data…)
Is there some way I can: A) prepare my EPIs such as to minimize the probability of an alignment error, and/or B) loop through and figure out the approximate shifts/angles I need to apply to each dataset…
…or am I: C) basically stuck nudging each dataset manually before running afni_proc.py?