Hi
You need to create such a file that contains the 90k grayordinates. To ensure
consistency we do that in MNI 2mm space:
- First create surface files in an FSL friendly format. Use the 32k surfaces,
as the vertices are ~2mm apart.
Left Surface:
surf2surf -i
$Subject/MNINonLinear/fsaverage_LR32k/${Subject}.L.white.32k_fs_LR.surf.gii -o
white.L.asc --outputtype=ASCII
--values=$Subject/MNINonLinear/fsaverage_LR32k//${Subject}.L.atlasroi.32k_fs_LR.shape.gii
Right Surface:
surf2surf -i
$Subject/MNINonLinear/fsaverage_LR32k/${Subject}.R.white.32k_fs_LR.surf.gii -o
white.R.asc --outputtype=ASCII
--values=$Subject/MNINonLinear/fsaverage_LR32k//${Subject}.R.atlasroi.32k_fs_LR.shape.gii
- Then extract the volume subcortical files from e.g.
$Subject/MNINonLinear/ROIs/Atlas_Rois.2.nii.gz, let’s say as
CIFTI_Structure_{Name}.nii.gz, using e.g. fslmaths or your favourite tool for
ROI extraction. You should get 19 of these NIFTI files. (You can obviously use
your favourite subcortical parcellation here, but if you want consistency with
the CIFTI standard grayordinates, you need to resample the final results, the
file above ensures consistency).
- Put all the filenames in a text file in the following sequence, to ensure
consistency with the rest of the CIFTIs:
white.L.asc
white.R.asc
CIFTI_STRUCTURE_ACCUMBENS_LEFT.nii.gz
CIFTI_STRUCTURE_ACCUMBENS_RIGHT.nii.gz
CIFTI_STRUCTURE_AMYGDALA_LEFT.nii.gz
CIFTI_STRUCTURE_AMYGDALA_RIGHT.nii.gz
CIFTI_STRUCTURE_BRAIN_STEM.nii.gz
CIFTI_STRUCTURE_CAUDATE_LEFT.nii.gz
CIFTI_STRUCTURE_CAUDATE_RIGHT.nii.gz
CIFTI_STRUCTURE_CEREBELLUM_LEFT.nii.gz
CIFTI_STRUCTURE_CEREBELLUM_RIGHT.nii.gz
CIFTI_STRUCTURE_DIENCEPHALON_VENTRAL_LEFT.nii.gz
CIFTI_STRUCTURE_DIENCEPHALON_VENTRAL_RIGHT.nii.gz
CIFTI_STRUCTURE_HIPPOCAMPUS_LEFT.nii.gz
CIFTI_STRUCTURE_HIPPOCAMPUS_RIGHT.nii.gz
CIFTI_STRUCTURE_PALLIDUM_LEFT.nii.gz
CIFTI_STRUCTURE_PALLIDUM_RIGHT.nii.gz
CIFTI_STRUCTURE_PUTAMEN_LEFT.nii.gz
CIFTI_STRUCTURE_PUTAMEN_RIGHT.nii.gz
CIFTI_STRUCTURE_THALAMUS_LEFT.nii.gz
CIFTI_STRUCTURE_THALAMUS_RIGHT.nii.gz
- You can then use such a file in FSL probtrackx2 as a seed (for Matrix1) or as
a target3 (for Matrix3). Notice that running a single probtrackx2 command with
all these seed locations and with many samples per location will require huge
processing power and memory. In practice, we run multiple probtrackx2 in
parallel, by splitting and parallelising the computation by using the --rseed
option in probtrackx2 and using a small number of samples --nsamples per
instance. I.e. instead of running one command, which will attempt to propagate
5000 curves per seed, we can run e.g. 100 instances of the above commands,
using --nsamples=50 and --rseed=i, i=1..100. Results can then be combined using
fdt_matrix_merge (which however also needs quite a lot of memory to produce the
final dense connectomes, but at least it is one final process).
- Another approach for parallelisation would be to split the seed locations and
merge the final matrices. There is no solution that fits all systems, you
should try and find the best approach depending on your computing resources.
Hope this helps
Stam
On 19 May 2015, at 01:01, David R. Haynor <[email protected]<mailto:[email protected]>>
wrote:
hi HCP,
we are trying to do tractography using FSL and HCP data. i have some questions
-- i suspect they have been answered already, but couldn't find those answers:
1. is there a file containing the grayordinate coordinates in the diffusion
space for a particular subject, both subcortical voxels and cortical vertices?
2. if i want to use some cortical and subcortical voxels from the list of
grayordinates as targets, do i have to run probtrackx2 twice (i.e. once for the
cortical surface vertices and once for the subcortical voxels), or can i run it
just once?
3. where is the label file for the grayordinate vertices/voxels for a given
subject?
thanks in advance.
-dh
_______________________________________________
HCP-Users mailing list
[email protected]<mailto:[email protected]>
http://lists.humanconnectome.org/mailman/listinfo/hcp-users
_______________________________________________
HCP-Users mailing list
[email protected]
http://lists.humanconnectome.org/mailman/listinfo/hcp-users
