You may be running into an unfortunate design problem in CIFTI-1, if NiBabel expects them to be standard NIfTI-2 - the dimensions are backwards in the header, so non-square matrices will have their elements rearranged. You may be able to get matlab to fix them via a reshape with the dimensions swapped. Alternatively, you can use wb_command -cifti-convert -to-gifti-ext to get them into a file that should read correctly per the GIFTI standard (and the matlab gifti toolbox).
CIFTI-2 will fix this problem, but hasn't been finalized yet (it is currently open for comments, see http://www.nitrc.org/forum/message.php?msg_id=9303). Tim On Mon, Jan 20, 2014 at 11:13 AM, Jason Bohland <jbohl...@bu.edu> wrote: > Hi. I've been working on putting together some ROI-level correlation > matrices from individual HCP data. I'm trying to use the wb_command > -cifti-parcellate to give me region level time series. I have been a bit > surprised by the results of this operation, and I wanted to verify that > what I'm seeing is correct, and ask a few questions. I appreciate any > guidance. I'll use subject 100307 as an example. > > 1) To get out cortical surface ROI time series, I started with the > following: > > wb_command -cifti-parcellate > rfMRI_REST1_LR_Atlas_hp2000_clean.dtseries.nii > ../../fsaverage_LR32k/100307.aparc.a2009s.32k_fs_LR.dlabel.nii COLUMN > 100307.cortical.ptseries.nii > > 2) To get out subcortical ROI time series, I used: > > wb_command -cifti-create-label subcortical_label.nii -volume > ../../ROIs/Atlas_ROIs.2.nii.gz ../../ROIs/Atlas_ROIs.2.nii.gz > > wb_command -cifti-parcellate > rfMRI_REST1_LR_Atlas_hp2000_clean.dtseries.nii subcortical_label.nii COLUMN > 100307.subcortical.ptseries.nii > > 3) Then I use Python with NiBabel ('1.4.0dev') to read and examine the > resulting matrices: > > ipython --pylab > > import nibabel as nb > > subcortical = > squeeze(nb.load('100307.subcortical.ptseries.nii').get_data()) > > subcortical.shape > Out[7]: (19, 1200) > > cortical = squeeze(nb.load('100307.cortical.ptseries.nii').get_data()) > > cortical.shape > Out[11]: (75, 1200) > > plot(subcortical.T) > hold(True) > plot(cortical.T) > xlabel('Scans') > > *** The resulting plot is attached *** > > Note the step-wise form of the curves, which does not appear correct (and > does not correspond well to the global mean in the dtseries file), > especially since the subcortical and cortical ROI curves are quite > different. In fact, it appears that the "number of steps" may equal the > number of ROIs (19 vs. 75) in each family of curves? This makes me think > that either nibabel is improperly loading the ptseries data, or > cifti-parcellate is not functioning correctly, but I haven't been able to > determine which is true. > > Is it possible to examine the data from the ptseries files any other way? > I tried -nifti-information -display-matrix but I get a message saying this > is not supported. I don't know of any other way to read in the ptseries > data. > > >>wb_command -nifti-information 100307.subcortical.ptseries.nii > -print-matrix > Printing Matrix/Volume information is only supported for Cifti. > > > As an additional question, assuming I'm going about this in a reasonable > way, and that I can find a way to read or export the resulting data, can I > assume that the ordering of the ROIs in the NIFTI/CIFTI matrix rows follows > the numerical ordering of the ids from FreeSurfer? > > Many thanks for your help, > -Jay > > > > > _______________________________________________ > HCP-Users mailing list > HCP-Users@humanconnectome.org > http://lists.humanconnectome.org/mailman/listinfo/hcp-users > _______________________________________________ HCP-Users mailing list HCP-Users@humanconnectome.org http://lists.humanconnectome.org/mailman/listinfo/hcp-users
