It sounds like SPM is the one re-orienting your data.

Those are GEMS I-files. I hope you are using either Imon or Dimon
to put the EPI timeseries data into AFNI datasets. That would be
too much of a pain directly using to3d. Or maybe you are using
Ifile...

In any case, those files should be well understood by to3d.

---

So you have acquired data in RSA orientation, which AFNI preserves.
However SPM probably expects it in either RPI or LPI orientation
(and if so, I don't know which).

So resample it in AFNI, before giving it to SPM:

3dresample -orient LPI -input 040902.RSA -prefix 040902.LPI.nii

Make *sure* the left and right are correct (since it will not
be obvious). Hopefully your ROI will help you do that.

- rick