Hi
Yes if you sum the average proportions (taken in the whole seed mask) across targets you should get a total of one.
The only case where this doesn't work is if there are voxels where zero samples get to any of the targets.
Consider a case where you have two targets, and say voxel i sends Xi samples to the first target and Yi samples to the second.
Imagine that we have already run proj_thresh, so Xi+Yi=1
Then averaging across the seed mask, you should get: sum(Xi)/N + sum(Yi)/N = sum(Xi+Yi)/N = 1
But if you have n voxels where no samples make it to any targets (Xi+Yi=0), then this total will be equal to (N-n)/N < 1
If instead of averaging across the whole seed mask, you only average non-zero voxels, then the number of such non-zero voxels will be different for different targets:
I.e. : sum(Xi)/M + sum(Yi)/N = sum(N.Xi+M.Yi)/NM != 1 in general
Cheers,
Saad.
On 26 Oct 2010, at 00:39, John Rize wrote:
> Hello,
>
> I have recently run a set of probtrackx using a single mask and several classification targets. I then used proj_thresh to threshold the output, then find the mean proportion of connection to each target. My question lies in the interpretation of these results. Since proj_thresh expresses the number of samples to each target as a proportion of total number of samples to any target, would one assume that summing the mean proportions of each target will be approximately equal to one? This is indeed the case for most of my subjects; there are however, a couple of subjects that these means sum to greater than one (i.e. ranging from 1.2 to 1.5). Am I incorrect in interpreting how proj_thresh works and do not need to be concerned about these values, or does this indicate an error in these couple of subjects?
>
> Thanks in advance for any help provided.
>
> J
>
--
Saad Jbabdi
University of Oxford, FMRIB Centre
JR Hospital, Headington, OX3 9DU, UK
(+44)1865-222466 (fax 717)
www.fmrib.ox.ac.uk/~saad
|