Hi
On 22 May 2009, at 08:54, SUBSCRIBE FSL Anonymous wrote:
> Hello,
>
> I 've read this paper: Beherens et al, Neuroimage 2007,34:144-55
> "we examin each fibre population to confirm that the volume fraction
> has not
> been forced to zero by automatic relevance determination (we
> threshold at
> 0.05)."
>
> Program(xfibres.cc) is written as
> "nfibs=0;
> if((m_sum_f[f]/m_nsamples>0.05)) nfibs++;
> ...
> m_beenhere = nfibs "
>
> But, mean_f2samples contains <0.05.
> What is the meaning of "we threshold at 0.05" ?
That paragraph in the paper refers to probtrackx, not xfibres (or
bedpotsx). It says that, when tracking through the fibre
distributions, only fibres with a mean f>.05 are considered.
The lines you extracted from xfibres.cc simply count, for that voxel,
how many fibres have survived the threshold of 0.05 (and this
information is used later on to initialise other neighbouring voxels).
> And m_beenhere is used by "bool neighbour_initialise".
> I donot understand "bool neighbour_initialise" well.
> I assume (possibly incorrectly) that it decides whether to
> execute "initialise_tensor".
neighbour_initialise is used to initialise the orientation and fibre
fraction of the second, third, etc, fibres in a voxel (instead of
using a tensor fit as initialisation).
>
>
> Many thanks for your help.
>
--
Saad Jbabdi
Oxford University FMRIB Centre
JR Hospital, Headington, OX3 9DU, UK
+44 (0) 1865 222545 (fax 717)
www.fmrib.ox.ac.uk/~saad
|