The bedpost directory itself remains untouched. I only transform
mean_fsamples and dyadic_vectors to standard space to be able to draw a
mask.
What I did to check if my matrix is right was:
- register mean_fsamples to template in standard space (flirt, 12 dof)
- apply this transformation to mean_fsamples in diffusion space to get
mean_fsamples in standard space
- invert transformation matrix
- apply inverted matrix to mean_fsamples in standard space, as this
would get me back to diffusion space
By comparing the original mean_fsamples in diffusion space with the
twice transformed mean_fsamples in diffusion space I tried to find out
how good the transformation really was.
Kerstin
----- Original Message -----
From: Matt <[log in to unmask]>
Date: Friday, July 6, 2007 12:01 pm
Subject: Re: [FSL] problem with probtrack
> Kerstin,
>
> If I understand correctly what you did, you should not apply
> transformationmatrices to the bedpost data. You should leave the
> bedpost outputs in
> diffusion space and use the transformation matrix as the input in
> probtrack.
>
> Regarding the masks, use avwmaths++ -in <maskfile> -div <maskfile> -
> out<binarymask file> to get only ones and zeros.
>
> Peace,
>
> Matt.
>
> -----Original Message-----
> From: FSL - FMRIB's Software Library [mailto:[log in to unmask]] On
> BehalfOf Ms Kerstin Hoffmann
> Sent: Thursday, July 05, 2007 9:49 PM
> To: [log in to unmask]
> Subject: Re: [FSL] problem with probtrack
>
> All masks are where I expect them to be, but the exclusionmask
> containsvalues of 0 and 0.9998... Running avwmaths <in> -bin <out>
> does not
> solve this problem.
> The path to the transformation matrix is fine, but applying the matrix
> to mean_fsamples in standard space and comparing with the original
> mean_fsamples shows some differences. The image histogram of the
> difference image of these 2 mean_fsamples has a FWHH of about .05 and
> 95% of all voxels have a difference below 0.15. Is this a difference
> that one would expect due to misregistration/interpolation?
>
> Thanks,
> Kerstin
>
>
>
> ----- Original Message -----
> From: Matt <[log in to unmask]>
> Date: Friday, July 6, 2007 10:55 am
> Subject: Re: [FSL] problem with probtrack
> > That syntax seems fine to me. You might verify that you can load
> > all the
> > masks (seed, targets, exclusion) you are calling in standard
> space in
> > FSLView and make sure they are where you expect them to be (and
> > have values
> > of only 1 or 0). Also make sure your path to the transformation
> > matrix is
> > correct (and that the matrix itself is good). My guess would be
> > that not
> > everything is in the proper space.
> >
> > Peace,
> >
> > Matt.
> >
> > -----Original Message-----
> > From: FSL - FMRIB's Software Library [mailto:[log in to unmask]]
> On
> > BehalfOf Ms Kerstin Hoffmann
> > Sent: Thursday, July 05, 2007 8:43 PM
> > To: [log in to unmask]
> > Subject: Re: [FSL] problem with probtrack
> >
> > Seed mask and target mask are in standard space, i am using the
> option> --xfm=<standard2diffsion-matrix>.
> >
> > probtrack.log contains this output:
> >
> > /usr/share/fsl/bin/probtrack --mode=seeds_to_targets -x
> >
>
/data/home/kerstin/data/stroke_recovery/alldti/Ker_Hof.01/proc/inclusionmask
> > .nii.gz
> > --forcedir -s
> >
>
/data/home/kerstin/data/stroke_recovery/alldti/Ker_Hof.01/bedpost.bedpost/me
> > rged
> > -m
> >
>
/data/home/kerstin/data/stroke_recovery/alldti/Ker_Hof.01/bedpost.bedpost/no
> > dif_brain_mask
> > --
> >
>
xfm=/data/home/kerstin/data/stroke_recovery/alldti/Ker_Hof.01/proc/mat/std2d
> iff.FA.mat
> > --
> >
>
rubbish=/data/home/kerstin/data/stroke_recovery/alldti/Ker_Hof.01/proc/exclu
> sionmask.nii.gz
> > -l -c 0.2 -S 2000 --steplength=0.5 -P 300
> > --
> >
>
targetmasks=/data/home/kerstin/data/stroke_recovery/alldti/Ker_Hof.01/proc/p
> robtrack/seeds2targets/targets.txt
> > --
> >
>
dir=/data/home/kerstin/data/stroke_recovery/alldti/Ker_Hof.01/proc/probtrack
> /seeds2targets
> >
> >
> > Kerstin
> >
> >
> > ----- Original Message -----
> > From: Matt <[log in to unmask]>
> > Date: Friday, July 6, 2007 10:31 am
> > Subject: Re: [FSL] problem with probtrack
> > > Can you post your probtrack.log output? Also, are you doing
> > > everything in
> > > native space?
> > >
> > > Peace,
> > >
> > > Matt.
> > >
> > > -----Original Message-----
> > > From: FSL - FMRIB's Software Library
> [mailto:[log in to unmask]]
> > On
> > > BehalfOf Ms Kerstin Hoffmann
> > > Sent: Thursday, July 05, 2007 8:29 PM
> > > To: [log in to unmask]
> > > Subject: Re: [FSL] problem with probtrack
> > >
> > > Yes, the diffusion directions are correctly orientated. Using
> > > probtrackwith seed mask option on my data gives results that
> are
> > > close to my
> > > expectations as well. The problem only occurs when using single
> seed> > voxel and connectivity based seed classification.
> > > Thanks for you help.
> > > Kerstin
> > >
> > >
> > > ----- Original Message -----
> > > From: Matt <[log in to unmask]>
> > > Date: Friday, July 6, 2007 10:06 am
> > > Subject: Re: [FSL] problem with probtrack
> > > > Did you verify that your diffusion directions are correctly
> > > > oriented with
> > > > respect to what you would expect anatomically (by loading the
> > FA
> > > > and V1 with
> > > > lines)?
> > > >
> > > > Peace,
> > > >
> > > > Matt.
> > > >
> > > > -----Original Message-----
> > > > From: FSL - FMRIB's Software Library
> > [mailto:[log in to unmask]]
> > > On
> > > > BehalfOf Ms Kerstin Hoffmann
> > > > Sent: Thursday, July 05, 2007 7:56 PM
> > > > To: [log in to unmask]
> > > > Subject: [FSL] problem with probtrack
> > > >
> > > > Hi!
> > > > I just ran into a problem using the connectivity based
> > > > seed-classification option in probtrack. Using all brain
> voxels
> > > (FA >
> > > > 0.1) as seed mask and an axial cut of the internal capsule as
> > target> > (number of samples 300), I got only about 15 voxels
> with
> > > intensity >0.
> > > > Using a seed mask inside the internal capsule and a target
> mask
> > > > right on
> > > > top of it returned equally poor results.
> > > > Also, when I use the single seed voxel option, the output
> file
> > > > containsno voxels with intensity >0, but using the same voxel
> > in
> > > > seed mask
> > > > option, I get results that look fine.
> > > > I also ran FEEDS, and no problems were reported.
> > > > Any advice would be greatly appreciated.
> > > > Thanks
> > > > Kerstin
> > > >
> > >
> >
>
|