To finally answer this, no, -wbsparse-merge-dense does not have any sparse summation logic, it is meant only for the dense (brain models) dimension, for concatenation. We don't currently have a command to do such a summation, trajectory support is somewhat a work in progress.
I don't know why running probtrackx multiple times would be significantly more memory efficient than increasing number of tracks per seed, you should probably ask on the FSL list. You also asked about dconn (cifti) files, which trajectory files are not. You can use -cifti-math on dconn files for summation (among other things), if that is really what you meant. Tim On Tue, Jun 24, 2014 at 3:48 PM, Colin Reveley <[email protected]> wrote: > Hi - > > when running FSL probtrackx with the "matrix 4" output, for use with > workbench, memory use is greatly (non-linearly) increased by sending out > more tracks per seed. The program is also much slower as memory use > increases like that. Memory use can be prohibitive, so we only seed a few > hundred lines which is not good statistically (hundreds of GB of RAM are > needed). > > it makes sense then to run the same task multiple times with different > random seeds, and then merge the results. > > I noticed -wbsparse-merge-dense but clearly this is meant for different > seed regions. It won't let me merge files where the surface seed and target > voxels are both "CORTEX_RIGHT" > > is there a way round this? would the merge operation actually be what I > want (a summation)? one could maybe rename the dimensions in the wbsparse > file if so? > > potentially the source FSL output files could be merged, then make the > traj file. I imagine I'd have to ask them about that, unless you happen to > know. > > to be clear I'm asking if it is possible to sum each count and distance > cell of dconn files to make a new one, only I want to do it with trajectory > files partly for display, partly they are very much smaller files. > > In future I'll parcellate the surface and merge the traj files seeded from > the parcels. Clearly that's the intent. > > best wishes, > > Colin Reveley > > _______________________________________________ > HCP-Users mailing list > [email protected] > http://lists.humanconnectome.org/mailman/listinfo/hcp-users > _______________________________________________ HCP-Users mailing list [email protected] http://lists.humanconnectome.org/mailman/listinfo/hcp-users
