Hi - we would advise very strongly against using 6 direction data in
bedpost - sorry - This is even more true if you only have one average
of your data.
The reason for this is that the model used in bedpost has 5 free
parameters + the variance of the noise.
Therefore, you effectively have 1 degree of freedom to estimate this
variance, so not only to do you get a very imprecise estimate of the
orientations, but also a very bad estimate of this noise variance.
The algorithm knows this, and therefore will give you huge
uncertainty on your orientations, so you will get very imprecise
tracking indeed - a very bad idea.
In general, it is not wise to acquire 6 direction DTI data, given how
quick it is to perform a useful DTI scan..
T
On 11 Dec 2007, at 17:57, Rajkumar wrote:
> Dear FSL Group,
> We have just started using FSL 4. So far, we were
> only doing
> deterministic tractography using DTI images (acquired using 6
> directional
> gradient with six repetitions is each direction). Recently we
> acquired some
> cases with 55 directions, and compared the probabilistic tract from
> both 6
> and 55 directions and what we found was that the tracts
> (Corticospinal tract,
> Arcuate fasciculus, optic radiation, and corpus callosum) with 55
> directions
> appeared bigger and more accurate than with 6 directions. Our
> questions are,
> 1) Can we use the six direction dti data sets to access the
> connectivity
> based seed classification of cortical and subcortical structures?
> 2) Does
> increasing the no of samples from 5000 to lets say 10000 help? (or
> Please
> provide other suggestions) 3) Are our 6 direction DTI data sets
> useless for
> probabilistic tractography? Thanks.
>
>
> Rajkumar.M.G
> Research Associate,
> PET Center,
> Children's Hospital of Michigan,
> Detroit.
>
|