Yes - absolutely as expected. There are 2 reasons -
1. some voxels have 2 fibres, but neighbours have 1, so the % of the
1st population will be different between them.
2. Even in the voxels with 2 voxels, in some voxels the each chain of
samples always stays with one population (e.g. f1 is always the same
population), but in other voxels f1 jumps between the two
populations, so "mean_f1" is a mixture of "f1 and f2". This obviously
has no effect on tractography as each sample is treated
independently, but it does make each image look a bit speckly. One
day we will get round to writing a program that unwraps the samples
so that f1 is always the same fibre population in each of the
samples, but as I say, it will make no difference to porbtrackx.
You can check all this by looking at the sum of mean_f1 and mean_f2.
This should be nice and smooth,
Cheers
T
On 19 Sep 2007, at 20:55, Timothy Laumann wrote:
> Dear FSL experts,
>
> We have been comparing the results of the new bedpostx to the old
> bedpost
> and have a question which is illustrated in the example jpeg I have
> attached (I hope it is large enough to see). Both sets of images
> are from
> the same subject with the same preprocessing. The left side
> represents the
> dyadic_vectors modulated by mean_fsamples from the original
> bedpost, the
> right side represents dyads1 modulated by mean_f1sample from bedpostx.
>
> We noticed that the bedpostx results have considerably more
> speckling in
> the image, i.e. scattered voxels which appear to represent a primary
> direction significantly different from the predominant direction which
> surrounds them. In this regard, the original bedpost image is much
> smoother. Is this speckling to be expected with the crossing fiber
> calculations or is this a result of the insufficient resolving
> power of
> our data acquisition scheme?
>
> Thanks so much for your help!
>
> Timothy Laumann
>
>
>
> <Example.jpg>
|