It is suggested from the FSL group to start your data analyses in an orientation most similar to what FSL's templates are stored.
this happens to be the "LAS - Radiological" orientation.
their software does not work well with our default orientation, "LPS - Neurological", which is why we have to use the initialization matrix to 'trick' their tools in the first place.
we do not recommend storing in "neurological" convention as not all programs will function correctly. We recommend that all data be converted to "radiological" convention before processed further with FSL.
Syam provided a very easy solution for this by using bxhorient, before converting your data into NIFTI format for processing:
bxhreorient --orientation=LAS $INDIR/YOUR.bxh $OUTDIR/temp.bxh
bxh2analyze -s --niigz $OUTDIR/temp.bxh $OUTDIR/data
I've been doing this for a while now, and its helped a lot with feat,first,fdt,etc processing. No need for any initialization matrices and some other problems with their anatomical analysis
tools have been cleared up as well.
syam has also mapped out what orientations should be fine.
> My impression was that FSL deals fine with the following orientations
> (which are strict rotations [no flips] of the LAS standard brain):
> Axial Sag Cor
> ----- ----- -----
> LAS ASL SLA
> ARS SPL LIA
> RPS PIL IRA
> PLS IAL RSA
> LAI ASR SLP
> ARI SPR LIP
> RPI PIR IRP
> PLI IAR RSP