Hi guys -
There is one slight Caveat to what Matt has said. If your data are
stored in neurological format, there is an inconsistency between
fslview and fdt, so even though the vectors will look correct in
fslview (either with dtifit or bedpost), your ttractography will
treat x as -x, and things will go wrong.
The safest way of solving this is to store all data radiologically.
Alternatively, you can set everything up so that it looks right in
fslview (with neurological data), and THEN negate the x values in bvecs.
We are changing fslview in the next release to make things consistent.
If your data are stored radiologically, Matt was absolutely right.
On 2 Aug 2007, at 02:00, Matt wrote:
> Assuming that you are giving DTIFit and BEDPOST the same data and
> bvecs you
> will get the same result for both programs. You can verify the
> directions in the BEDPOST by loading the dyadic vectors (or
> similarly named
> file) the same way as V1 and mean_fsamples the same as FA.
> -----Original Message-----
> From: FSL - FMRIB's Software Library [mailto:[log in to unmask]] On
> Of Ruth Carper
> Sent: Wednesday, August 01, 2007 7:55 PM
> To: [log in to unmask]
> Subject: [FSL] mismatch between DTIFit and BEDPOST results
> weeell, It's clear I've got some sort of issue with getting my
> right. I've got the right vectors, but I wasn't sure about the
> signs. So I
> tried about every permutation of negating x,y, and/or z in my bvecs
> file and
> running it through DTIFit. Looking at the primary eigenvectors
> from each permutation made it pretty obvious which was the right
> there was only one result that gave V1 line segments (in FSL_View)
> matched the corpus callosum anatomy in all directions. However,
> when I ran
> those exact same data and bvecs through BEDPOST, it was as if it
> was using a
> bvecs file where x had been negated (x*-1). Even if I do have an
> error in my
> bvecs file, shouldn't I get the same erroneous result from both
> I'll rerun BEDPOST tonite to be sure it replicates, but I already
> checked all my command lines.