Ha - that would be good.
No, I'm afraid you've got classic 12 direction data.
FDT will work with this dataset, you are right. And it is likely to
give perfectly good results (see Heiervang et al. 2007 Neuroimage for
quantitative comparison between 12 and 60).
Historically, the reason we have suggested 25 is that with any fewer
directions the expected uncertainty is dependent on the fibre
orientation. For example, depending on your gradient orientations,
you may get a bias towards more spread in the z direction etc. (See
Jones et al paper on uncertainty and sampling schemes)
But you can certainly try with 12.
Cheers
T
On 8 Mar 2007, at 22:35, Peter Smale wrote:
> Hi Team FSL,
>
> Regarding the discussion about DTI directions:
>
> My main question is: when you say a DIRECTION, do you mean a vector
> pointing
> in one direction, or do you mean a line that actually points in two,
> opposite directions?
>
> I have been wondering because I have been doing some DTI of stroke
> patients
> using diffusion data from a 1+12 direction Siemens pulse sequence.
> I read on
> the help pages that at least 25 directions should be used for
> probtrack.
>
> I fed in my 12 directions anyway and they have yielded some very
> reasonable
> looking results. Similar results to Cherubini et al in NeuroImage (Jan
> 2007), who also used only 12 directions on a Siemens scanner. I
> wonder if
> this is because what Siemens calls a 1+12 direction scan actually
> has 25
> directions: 12 directions, plus their opposites, plus the b = 0
> scan. This
> assumption is based on the way their directions are calculated. For
> example,
> the built-in set of 30 directions is calculated from the structure
> of a
> bucky-ball which has 60 points, the 10 direction set is found from the
> vertices of a dodecahedron (20 points).
>
> Is the FSL definition of direction the same? i.e. Is 25 directions
> the same
> as 25 DWI volumes, or is it actually 12 volumes plus 1 b= 0 volume?
> Am I
> alone in my monotonous definition of direction?
>
> Thanks in advance,
> Peter
>
|