Print

Print


deterministic tractograohy is nice to get an overview of the data, although
bedpostx/probtrackx tends to be better for everything else in general
(unless your research is in diffusion imaging)

In both cases it's wise to carefully consider what the data actually
represents. which is documented.

dti tracking or qball if you have the data can be processed and tracked
determisitically in pretty 3d etc using trackvis. you don't need to worry
about voxel ordering etc anymore. it's interoperable with raw fsl 4d data
-++.

just transpose the bvecs for trackvis, copy, paste into trackvis gui and
thats it. probably.

you can even apply a flirt .mat linear transform to the trackvis tracks.
docs @ fsl and trackvis site.

if you have 60+ gradients you could try qball in trackvis and bedpostx in
fsl. odf a *tad* more involved than tensors in trackvis. but making tensors
+ tracks v easy.

you can use the masks that you make for fsl as roi in trackvis without
alteration. use the binary  no-dif-brain mask as 2nd threshold volume for
trackvis dti/odf recon.

you'd be way better off with trackvis than misusing fsl. and although they
don't interoperate they are on speaking terms.

Colin


On 20 Apr 2013, at 09:29, Gabor Perlaki <[log in to unmask]> wrote:

> Dear all,
>
> Can I use streamline tractography on diffusion data generated with FSL's
dtifit or is it any program capable to use
>
> <Bassname>_V1 and <Bassname>_FA data for streamline tractography.
>
>
> Thanks,
> Gabor
>

Date:    Sun, 21 Apr 2013 09:44:20 +0100
From:    Saad Jbabdi <[log in to unmask]>
Subject: Re: FSL streamline tractography

Hi
You can use probtrackx for deterministic tracking, e.g. if you use
orientation distributions that have only one sample. You will need to
convert your V1 orientation file to theta/phi angles.
Note however that probtrackx is not designed to do deterministic
tractography. Using probtrackx this way is sub-optimal for two reasons: (1)
The results are stored as spatial histogram counts rather than individual
streamlines; and (2) probtrackx performs "probabilistic interpolation".
If you still want to use probtrackx, then you can use fslroi/fslmaths to
transform V1 to theta/phi by inverting these transformations:

V1x = sin(theta)cos(phi)
V1y = sin(theta)sin(phi)
V1z=cos(theta)

Cheers
Saad