Hi
> On 21 Dec 2016, at 10:05, Elisabetta Giacalone <[log in to unmask]> wrote:
>
> Hi FSL community,
> I'm trying to obtain a connectivity matrix from a whole brain tractography using probtrackx. I used the --network option and I obtained a fdt_network_matrix 82x82 using cortical and subcortical volumes (from freesurfer parcellation) . As I read in this mailing list, I normalized it dividing each row of the matrix by the corrisponding waytotal.
> I have some questions:
>
> 1) Is it correct to normalize by waytotal or should I do it by the sum/average of volumes of the connecting regions too?
No, what you did (normalise each row by its corresponding waytotal) was correct.
>
> 2) I tried to correct the path distribution by the distance between regions using the --pd option (mantaining the --network) but I verified that the fdt_network_matrix and waytotal remain the same as them without correction. Is it correct?
Yes the network matrix that is saved only stores the number of samples, it does not multiply them by the distance even when you turn on the -pd flag.
>
> 3) How can I obtain a connectivity matrix corrected by the distance between regions?
This is not straightforward. You can track from each seed roi separately and use the —ompl flag to get the distances to each of the other rois. (which you can extract from the output distance file using fslstats). However bear in mind that this distance correction is just a heuristic and is not meant to be used in a quantitative way.
>
> 4) finally, my fdt_network matrix is not symmetric. In order to analyze it, after normalization, is it correct to symmetrise it averaging the symmetrical elements?
Yes. The interpretation then is a little different. Instead of storing the probability that a streamline from A goes to B, it will give you the probability that a streamline from either A or B goes to B or A. Of course neither of these things is a direct measure of connection strength, as we have always stressed :)
Cheers
Saad
>
> I attach my probtrackx.log, fdt_network_matrix and the waytotal file.
>
> Thank you a lot.
> Elisabetta
|