Print

Print
Hi Saad,

Thanks for the response!  I actually found something different when I tried running an experiment with two cases: theta = [0,pi) and theta = arccos(2v-1).  Phi was the same in both of these cases: phi = [0,2i).  I have attached the results.  Green is the distribution with the former and red with the latter.  Without the arccos modification, the fibers mainly course in the direction of the poles, as one would expect when switching from Spherical to Cartesian coordinates.  With the arccos modification, things appear spherically uniform in Cartesian.  Thus it seems that spherical distributions result when arccos is used.  What are your thoughts on these results?  Thanks for your help!

-Kabi

On Sat, Dec 12, 2015 at 6:14 AM, Saad Jbabdi <[log in to unmask]> wrote:
Hi - you need to use a uniform sampling on the sphere using theta and phi as defined here: 
http://fsl.fmrib.ox.ac.uk/fsl/fslwiki/FDT/UserGuide?action=AttachFile&do=view&target=fdt_spherical_polars.gif

Cheers,
Saad


On 11 Dec 2015, at 21:42, Kabilar Gunalan <[log in to unmask]> wrote:

Hi Saad,

Thanks for the response!  Just to clarify, do the diffusion directions need to be uniformly distributed in the spherical or Cartesian coordinate system?  I chose phi and theta so that when transformed to Cartesian coordinates, the direction vectors were uniformly distributed.  That is, theta = arccos(2v-1) and phi = [0,2i), where v is randomly sampled from (0,1).

Thanks!

-Kabi

On Thu, Dec 10, 2015 at 6:08 AM, Saad Jbabdi <[log in to unmask]> wrote:
Hi

Yes, you can generate random theta/phi using uniform distribution on the sphere, then set all the volume fractions to 1 (not zero).

cheers
Saad


On 9 Dec 2015, at 20:55, Kabilar Gunalan <[log in to unmask]> wrote:

Hi all,

How do I run probtrackx(2) with an isotropic fiber orientation distribution?  Do I set all volume fractions (i.e., in merged_f*samples.nii.gz) to zero, or do I also need to modify merged_ph*... and merged_th* so that the sampled diffusion directions are uniformly distributed around the unit sphere?

Based on some preliminary runs, it appears merged_ph*... and merged_th*... must be modified, along with setting all volume fractions to zero.  I chose phi and theta so that when transformed to Cartesian coordinates, the direction vectors were uniformly distributed.  That is, theta = acos(2v-1) and phi = [0,2i), where v is randomly sampled from (0,1).

Thanks for all of your help!!

-Kabi

------------------------------------------------------------------
Saad Jbabdi, PhD
Associate Professor
MRC Career Development Fellow

FMRIB Centre, University of Oxford,
John Radcliffe Hospital, Oxford, OX3 9DU, UK. 
tel (+44)1865-222466  (fax 717)






------------------------------------------------------------------
Saad Jbabdi, PhD
Associate Professor
MRC Career Development Fellow

FMRIB Centre, University of Oxford,
John Radcliffe Hospital, Oxford, OX3 9DU, UK. 
tel (+44)1865-222466  (fax 717)