Dear Ed,
I notice that Steve responded to your original
email, but you didn't seem to acknowledge
that - as you sent a repeat email later that
same day. His reply was:
"Hi - sounds like the problems are bad enough that there's not much
point looking at these. Did you have all brains in the same
orientation as the MNI152 to start with?
If not, you could fix all those with fslswapdim before beginning TBSS."
Have you done this?
There is a tool in FSL 4.1.6 (which I would
recommend you use) that allows you to
easily perform this operation. It is called
fslreorient2std. It assumes that the labels
are correct in FSLView (and I assume they
are because you shouldn't have done any
analysis at all if they were not).
So do this, or let us know if you have done it
and it still does not work. Although you talk
about using fslswapdim in your email, it is
unclear whether you are using it in this way or
not. Just flipping one axis with fslswapdim
will cause a problem with the left/right
orientation, so you should not do this. I
hope this is not what you have done. If
so, please let us know and give us full
details.
All the best,
Mark
On 1 Sep 2010, at 10:16, Ed Gronenschild wrote:
> Hi,
>
> I posted this before on 4 July and recently
> on 24 August.
>
> We are experiencing serious problems with
> TBSS. A close inspection of the 4D stack
> all_FA.nii.gz in the stats folder revealed
> that more than 80 % (!!!!) of the FA maps
> were badly registered to the default FA
> template FMRIB58_FA_1mm (we used the -T
> option in tbss_2_reg).
> It turned out that these problems got
> resolved by flipping the y coordinate of
> the offending FA maps beforehand (using
> fslswapdim), although these were correctly
> oriented (as shown by fslview).
> Our conclusion is that by some unknown
> reason the registration tool fsl_reg
> spontaneously introduces a y flip
> of the data in the majority of the cases.
>
> These problems persisted in three different
> projects, two of which based on a Siemens
> MRI scanner and one based on a Philips
> scanner.
>
> We used FSL version 4.1.4.
>
> Ed
>
|