-cifti-separate and -cifti-convert have very different purposes.
-cifti-separate is there to make usable volume and metric files from the
information in cifti files.  -cifti-convert is there only to get data
values into other tools that can't read cifti, without any respect for
spatial information.  Thus, -cifti-separate has to add in all the excluded
voxels and vertices to its outputs, while -cifti-convert does not (however,
-cifti-convert -to-nifti does have to work around the 32,767 limit on
nifti-1 dimensions, which makes things interesting).

The current commands and the 91k grayordinates space always have the
vertices in ascending order.  The 0 in your example is just a placeholder
for missing data (medial wall), not a change in order.  It is possible to
make a cifti file with arbitrary order of vertices within a hemisphere (you
can't interleave across hemispheres or structures in general), but
wb_command doesn't have a way to make such a file from scratch (it will
accept such files, if anyone makes them, but it strikes me as undesirable,
to the point of thinking maybe the specification should have dictated the

The -cifti-export-dense-mapping command will give you the index translation
information for any dense cifti file.  We may have released some cifti
files that did not exclude the medial wall, and we have released some that
have only surface data, and no volume data (~59k indices).  All standard
91k grayordinate cifti files have the same index mapping (which is the
definition of what makes them standard).


On Mon, Jul 2, 2018 at 5:27 PM, Jo Etzel <jet...@wustl.edu> wrote:

> Is there a way to determine (or specify) the correspondence between the
> vertices when a cifti is converted to a "fake nifti" vs. separated to
> giftis?
> For example, wb_command -cifti-separate Parcels_LR.dtseries.nii  COLUMN
> -metric CORTEX_LEFT out.func.gii produces a gifti with 32492 vertices,
> the first 10 of which have the values 1 2 3 4 5 6 7 0 8 9.
> Converting the same cifti via wb_command -cifti-convert -to-nifti
> Parcels_LR.dtseries.nii out.nii.gz -smaller-file gives a file of
> dimension 29706, 2, 1, 1. 29706*2 = 59412, which is also the number of
> rows in the text file that is given by wb_command -cifti-convert
> -to-text Parcels_LR.dtseries.nii out.txt. The first 10 values of these
> files is 1  2  3  4  5  6  7  8  9 10, different than the gifti.
> My guess is that the -cifti-convert output has fewer vertices than the
> -cifti-separate output because the medial wall is omitted. But is there
> a fixed ordering to how the 29706 cifti vertices fit into the 32492
> gifti vertices? (Rephrased, is it something like the vertex in row
> 17,234 of the -cifti-convert -to-text file is always assigned to slot
> 10,222 of the -cifti-separate gifti file?) If the reassignment is fixed,
> is it available somewhere?
> thanks,
> Jo
> --
> Joset A. Etzel, Ph.D.
> Staff Scientist
> Cognitive Control & Psychopathology Lab
> Washington University in St. Louis
> mvpa.blogspot.com
> _______________________________________________
> HCP-Users mailing list
> HCP-Users@humanconnectome.org
> http://lists.humanconnectome.org/mailman/listinfo/hcp-users

HCP-Users mailing list

Reply via email to