Missing nodes in parcellation image

1

Hi there,

I am using the following code to

  1. map the annotation files of the Glasser atlas from fsaverage to the subject
  2. map the Glasser annotations onto the volumetric image and add (Freesurfer-specific) subcortical segmentation
  3. Replace the random integers of the Glasser file with integers that start at 1 and increase by 1
  4. register the ordered atlas-based volumetric parcellation to diffusion space
  5. generate connectome
      
      #map the annotation files of the HCP MM1.0 atlas from fsaverage to the subject (first left, then right hemisphere)
      mri_surf2surf --srcsubject fsaverage --trgsubject $subjectID --hemi lh --sval-annot $SUBJECTS_DIR/fsaverage/label/lh.hcpmmp1.annot --tval $subjectDIR/label/lh.hcpmmp1.annot
      mri_surf2surf --srcsubject fsaverage --trgsubject $subjectID --hemi rh --sval-annot $SUBJECTS_DIR/fsaverage/label/rh.hcpmmp1.annot --tval $subjectDIR/label/rh.hcpmmp1.annot
      
      #Map the HCP MMP 1.0 annotations onto the volumetric image and add (Freesurfer-specific) subcortical segmentation.
      #Convert the resulting file to .mif format (use datatype uint32 --> liked best by mrtrix)
      mri_aparc2aseg --old-ribbon --s $subjectID --annot hcpmmp1 --o $derivatives_path"/dwi/"$sub"/"$ses"/anat/"$subjectID"_hcpmmp1.mgz"
      mrconvert --datatype uint32 $derivatives_path"/dwi/"$sub"/"$ses"/anat/"$subjectID"_hcpmmp1.mgz" $derivatives_path"/dwi/"$sub"/"$ses"/anat/"$subjectID"_hcpmmp1.mif"
      
      #Replace the random integers of the hcpmmp1.mif file with integers that start at 1 and increase by 1
      labelconvert $derivatives_path"/dwi/"$sub"/"$ses"/anat/"$subjectID"_hcpmmp1.mif" /home/julia/mrtrix3/share/mrtrix3/labelconvert/hcpmmp1_original.txt /home/julia/mrtrix3/share/mrtrix3/labelconvert/hcpmmp1_ordered.txt $derivatives_path"/dwi/"$sub"/"$ses"/anat/"$subjectID"_hcpmmp1_parcels_nocoreg.mif"

      Register the ordered atlas-based volumetric parcellation to diffusion space
      first calculate transformation
     
      flirt -in $derivatives_path"/dwi_oct2020/"$sub"/"$ses"/dwi/"$subjectID"_dir-APPA_meanB0brain.nii.gz" -ref $derivatives_path"/dwi_oct2020/"$sub"/"$ses"/anat/"$subjectID"_acq-mprage_T1wBrain.nii.gz" -dof 6 -omat $derivatives_path"/dwi_oct2020/"$sub"/"$ses"/dwi/"$subjectID"_diff2struct_fsl.mat"
      transformconvert -force $derivatives_path"/dwi_oct2020/"$sub"/"$ses"/dwi/"$subjectID"_diff2struct_fsl.mat" $derivatives_path"/dwi_oct2020/"$sub"/"$ses"/dwi/"$subjectID"_dir-APPA_meanB0brain.nii.gz" $derivatives_path"/dwi_oct2020/"$sub"/"$ses"/anat/"$subjectID"_acq-mprage_T1wBrain.nii.gz" flirt_import $derivatives_path"/dwi_oct2020/"$sub"/"$ses"/dwi/"$subjectID"_diff2struct_mrtrix.txt"
      mrtransform -force $derivatives_path"/dwi/"$sub"/"$ses"/anat/"$subjectID"_hcpmmp1_parcels_nocoreg.mif" -linear $derivatives_path"/dwi_oct2020/"$sub"/"$ses"/dwi/"$subjectID"_diff2struct_mrtrix.txt" -inverse -datatype uint32 $derivatives_path"/dwi/"$sub"/"$ses"/anat/"$subjectID"_hcpmmp1_parcels_coreg.mif"
      mrstats $derivatives_path"/dwi/"$sub"/"$ses"/anat/"$subjectID"_hcpmmp1_parcels_nocoreg.mif"
      mrstats $derivatives_path"/dwi/"$sub"/"$ses"/anat/"$subjectID"_hcpmmp1_parcels_coreg.mif"

      #connectomics
      connectome_filename=$derivatives_path"/dwi/"$sub"/"$ses"/dwi/"$subjectID"_connectome_Glasser.csv"
      tck2connectome $derivatives_path"/dwi/"$sub"/"$ses"/dwi/"$subjectID"_iFOD2.tck" $derivatives_path"/dwi/"$sub"/"$ses"/anat/"$subjectID"_hcpmmp1_parcels_coreg.mif" $connectome_filename --assignment_radial_search 3.2 -zero_diagonal --force

For some sessions, I am getting the following warning (with changing node numbers):
tck2connectome: [WARNING] The following nodes are missing from the parcellation image:
tck2connectome: [WARNING] 120
tck2connectome: [WARNING] (This may indicate poor parcellation image preparation, use of incorrect or incomplete LUT file(s) in labelconvert, or very poor registration)

I checked the parcellation and it’s fine. What could be causing this?
thanks in advance for any help!
Julia

2

Welcome Julia!

  1. Replace the random integers of the Glasser file with integers that start at 1 and increase by 1

Well, not truly “random” I hope; that would make performing the reordering slightly difficult :sweat_smile: Let’s call them “non-sequential”.

For some sessions, I am getting the following warning (with changing node numbers):
I checked the parcellation and it’s fine.

Did you confirm that there is at least one voxel containing the value 120 in the parcellation image? tck2connectome is claiming that no such voxel exists. labelstats can tell you the number of voxels in the image that possess each integer label present.

I must admit I sometimes giggle a little when people ask about this particular warning, given that the verbosity of the warning message is quite a lot larger than many others. But it’s supposed to be a very general “there might be something wrong with your data; do a quality check” type message. So if tck2connectome claims there are no such voxels, then work your way up the pipeline from bottom to top, checking to see if you can find any such voxels at any point. At the labelconvert step, there must be some integer label in the image provided at its input that is supposed to map to 120 in its output image, so check for the presence of that value in the input image. If it’s not there either, then check the outputs of the FreeSurfer commands to see whether any such voxel is generated at any point. If it is there, but labelconvert makes it disappear, then you’ve identified the point of failure. Given that the observation of a potential fault is flagged at the end of a pipeline, all you can do is crawl through the data slowly; if I knew from the warning message what the cause is, I’d have changed the warning message to make it say so :stuck_out_tongue:

Good luck!
Rob