Dear Tim,
Thanks for the follow-up. Could you suggest a quick solution for this that can
do the work until the release of the fixed data file? I am not much familiar
with most of the wb_command,- as you can see in my original post, I am more of
a Matlab person- so perhaps, if you could give me a hint about how to change
those label files, I would really appreciate. A command that can be used for
filling the holes in the 4th map in order to regenerate the 3rd map would also
help.
Best regards,Salim
From: Timothy Coalson <[email protected]>
To: Salim Arslan <[email protected]>
Cc: "[email protected]" <[email protected]>
Sent: Friday, August 5, 2016 12:46 AM
Subject: Re: [HCP-Users] Extracting Power parcellations from
RSN-networks.32k_fs_LR
Using -cifti-separate on the original cifti file and viewing in wb_view looks
identical to the original. However, I think I see what is going on, after
getting a hint from the way you reset the color scheme - the file has 2 labels
named "a3_Default_mode", and similar for other labels, and turning one of them
off reveals these holes.
What has happened is that someone has set the names and colors of two label
values to be identical (except for a space on the end of one, to slip past the
uniqueness tests on the label names), but they are actually different integers
in those patches. I don't know how this happened, as map 4 doesn't have this
difference in values. We should probably fix this in the data file.
Tim
On Thu, Aug 4, 2016 at 3:44 AM, Salim Arslan <[email protected]> wrote:
Hi Tim,
Thanks for your attention.
After your message, I also thought the problem was about the order of the
hemisphere in the cifti file. However, after running wb_command
-file-information -no-map-info I can confirm that the order is correct. Here is
the output:
Name: RSN-networks.32k_fs_LR. dlabel.niiType:
Connectivity - Dense LabelStructure: CortexLeft CortexRight
Data Size: 1.04 MegabytesMaps to Surface: trueMaps to
Volume: falseMaps with LabelTable: trueMaps with Palette:
falseNumber of Rows: 64984Number of Columns: 4Volume Dim[0]:
0Volume Dim[1]: 0Volume Dim[2]: 0Palette Type:
NoneCIFTI Dim[0]: 4CIFTI Dim[1]: 64984ALONG_ROW map
type: LABELSALONG_COLUMN map type: BRAIN_MODELS Has Volume Data:
false CortexLeft: 32492 out of 32492 vertices CortexRight:
32492 out of 32492 vertices
I do not think anotherCii has also nothing to do with this, as it is just a cii
file with the same row dimensionality (64984x1).
Following the suggestion in your other message, I have extracted the
hemisphere-specific full-surface label files using wb_command -cifti-separate
with the -label repeatable option and saved the output in a gii file. The
result was the same.
One thing I forgot to mention and might be important is that, among 4
parcellations stored in RSN-networks.32k_fs_LR.dlbal. nii, only the
holes-filled power labels (3rd column) has this problem. The others are just
fine. This also makes me think that dlabel file, at least partly, was
generated in an unusual way. Perhaps someone involved in the generation of the
tutorial data package or the aforementioned BALSA study can shed light on this.
Salim
From: Timothy Coalson <[email protected]>
To: Salim Arslan <[email protected]>
Cc: "[email protected] " <[email protected] >
Sent: Wednesday, August 3, 2016 9:36 PM
Subject: Re: [HCP-Users] Extracting Power parcellations from
RSN-networks.32k_fs_LR
First, the size of 64984 shows that the cifti file in question did not have the
medial wall masked out internally, which is likely the starting point for these
problems. I don't know how medialMask is defined, but if it isn't 32492
elements long (in this case), then that will cause another problem, as the
break between left and right data will be in the wrong place.
If you really didn't do any steps between defining labelsL and labelsR, and
anotherCii was made as a copy of that same cii variable, then I don't know how
it managed to mess things up, though, unless this cifti file also managed to
have right hemisphere before left hemisphere (the format supports this, but we
have tried to be consistent with the order - the lack of a medial wall mask
indicates this dlabel file was generated in an unusual way, though). You can
use "wb_command -file-information -no-map-info" to see the order of structures
in the cifti file.
Tim
On Wed, Aug 3, 2016 at 11:22 AM, Salim Arslan <[email protected]> wrote:
Hello,
I am trying to get Power parcellations out of RSN-networks.32k_fs_LR.dlabel.
nii, that comes with the "Brain Analysis Library of Spatial maps and Atlases
(BALSA) database study" (https://balsa.wustl.edu/ study/show/WG33). It is also
part of the tutorial dataset.
I did what I have been always doing for extracting data from CIFTI files in
Matlab. Using ciftiopen, I obtained the cii object, which in this case,
contains the cdata matrix of size 64984x4. The 3rd column here corresponds to
the holes-filled Power parcellations. To extract those for both hemispheres, I
simply split up the 64984 vector and masked out the medial wall vertices and
then saved the labels back to CIFTI.
cii = ciftiopen('RSN-networks.32k_ fs_LR.dlabel.nii', wb_command);
cdata = cii.cdata;
parcellation = cdata(:,3);
labelsL = cifti(1:length(medialMask));
labelsR = cifti(length(medialMask)+1: end);
anotherCii.cdata = [labelsL;labelsR];
ciftisave(anotherCii,'power. dlabel.nii',wb_command)
However when I visualized the labels on workbench, I noticed some holes (with
random labels) in many parcels. Please see the attached figure for
clarification.
I assume this is something to do with the medial masking, but cannot figure out
what is wrong. Any help is appreciated.
Thanks for your attention.
Best regards,Salim
--
______________________________ _________________
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
