L.S.,
On this mailing list I found a reply by Saad Jbabdi (https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=FSL;d5138668.1103) describing the use of the --omatrix1 flag for probtrackx to get voxel-to-voxel connectivity. This was exactly what I needed, so great news :-)
However, I'm running into some strange behaviour. The command I use is:
${FSLDIR}/bin/probtrackx --mode=seedmask -x $WMM -s $s -m $nodif_mask -l -S 2000 -P 100 -c 0.2 --forcedir --dir=$out --omatrix1
The seedmask $WMM consists of a selection of isolated voxels on the white-grey matter boundary. Everything is in diffusion space. The low sample size is for testing purposes only.
When the process is complete I get an fdt_ matrix1 file of correct dimensions, and a correct matrix for the voxel-coordinates, but when I view the connectivity matrix in Matlab I get this picture: http://cs.ru.nl/~mhinne/tmp/omatrix1.jpg. This is clearly not the true connectivity. Also, increasing sample size does not change this matrix, which should reflect the number of accepted streamlines, right?
Can someone spot the error in my approach, or is --omatrix1 not intended to be used this way? I'm using FSL 4.1.7.
Also, when combining --omatrix1 with --opd, the fdt_paths matrix is empty. Removing the --omatrix1 flag then results in correct fdt_paths.
|