I think you are right - we do not wrap the angles in bedpostx - if
you want them wrapped, you can do angle mod 2 pi, as you suggest.
Cheers
T
On 30 Oct 2007, at 08:23, Jeroen Siero wrote:
> Dear all,
>
> I am doing a small experiment (comparing different acquisition
> approaches
> and averaging of datasets) using the distributions of the angles
> theta and
> phi obtained from the BEDPOSTx procedure (using only on fibre). I
> tried
> using FSL 3 (using the 'old' BEDPOST) and I observed that the phi
> angles
> always range between 0 and 2*pi, and theta between 0 and pi
> (looking at the
> histograms in fslview). However, when using FSL4.01 with BEDPOSTx the
> histograms in the merged_th1_samples and merged_ph1_samples (for a
> particular volume) suddenly appear a bit strange, the values range
> from
> approximately -1e11 and 1e11 with a high peak around zero. Most of the
> voxels are in the order of a fews radians, what you would expect
> from the
> ranges of theta and phi, but some have very strange high values.
> The merged
> files I looked at are the ones straight from the BEDPOSTx output (no
> averaging has been done or other manipulations). I was wondering
> whether in
> FSL4.01 the priors on theta and phi used in the Bayesian estimation
> procedure are bounded to the range of 0 and 2pi resp. 0 and pi or
> that the
> angles can 'rotate' infinitely on the unit sphere (hence I should do
> something like angle mod 2pi).
> Could you explain why I get these strange histograms? (it happens
> in all of
> the 10 subjects).
> Thank you for your help!
>
>
> Best regards,
>
> Jeroen Siero
> BSc. Applied Physics
> Master student Biomedical Engineering
> NeuroImaging Center Grongingen, The Netherlands
|