Hi Anthony,
If both the 'align' and 'tlrc' blocks are included,
there should be very little difference between e2a
and a2e.
But e2a seems to make more sense in that it goes the
direction of anatomical accuracy (the T1 images tend
to have much less distortion than T2* images). The
standard 'align' transformation is affine though, so
only scalar distortions would be applied anyway.
For those wanting to process data in orig space, it
is common to stick with the default a2e (i.e. do not
add any e2a option). That keeps the EPI data as
untouched as possible.
But particularly when going to standard space, e2a
seems a little more robust.
- rick