Dear MRtrixters,
I am working with HCP data, and have generated two parcellation images along with a 5TT (hsvs) image. All three images have been generated using the same FreeSurfer outputs. The parcellation images include:
379 node HCPMMP1 atlas (as per the BATMAN tutorial)
I have noticed that the cortical ribbon in the DK parcellation image matches up with the cortical ribbon in the 5TT image (attached, left figure). However, the cortical ribbon in the HCPMMP1 image has a few (possibly problematic) differences compared to the 5TT image (right figure, see crosshair).
Any ideas on why the ribbon thickness is different between the two parcellation images, and, how to generate a HCPMMP1 parcellation image with the same ribbon thickness as the 5TT image?
I’m not sure why this would happen although it looks like a local rather than a global problem - seems like the label is completely missing there so perhaps worth debugging the steps in the creation of the parcellation.
In the past, I’ve used the script from here to obtain satisfactory HCPMMP1 parcellations after running recon-all.
Hi Nick,
Thanks for getting back to me so soon - your recommendation worked well
The next step is to re-index the parcellation file to have nodes incrementing from 1. I am unsure how to address this with the available outputs - in particular, LUT_HCPMMP1.txt. This lookup table doesn’t seem to match up with the values in HCPMMP1.nii.gz (for example, node 28 is present in HCPMMP1.nii.gz but is absent from LUT_HCPMMP1.txt).
I have pasted the contents of LUT_HCPMMP1.txt below - is this a sensible output, or has something gone wrong?
Also, do you know where I can access the lut_in and lut_out files for labelconvert?
It can happen that a participant misses a certain parcel - depending on their anatomy - so don’t use the participant LUTs from the output of that script, as they will only contain the labels present in that participant.
You can find the files for labelconvert here. hcpmmp1_original is the reference LUT and hcpmmp1_ordered gives the desired order for the output.
It has been brought to my attention that node 2001 (R_V1) has been mislabelled.
Here are a few screenshots showing the error. The first screenshot shows the parcellation image generated using FastSurfer inputs, while the second shows the parcellation image generated using FreeSurfer inputs. In each screenshot, node 2001 is displayed as a region of interest, and the main image is hcpmmp1.nii.gz (created from the script create_subj_volume_parcellation.sh).
The command used to generate these images: bash ${ResultsDir}create_subj_volume_parcellation.sh -L ${ResultsDir}FreeSurfer_DirName.txt -a hcpmmp1 -d ${sub}_HCPMMP1_parcellation
Has this been previously identified, and is there a fix? Happy to share intermediate images (i.e., FastSurfer/FreeSurfer outputs) if it would be helpful.
Label 2001 in the lookup table is indicated as unknown, while label 2002 is the R_V1. In my FreeSurfer output (I haven’t used FastSurfer), the R_V1 label seems to be in the right place.
Could you check whether label index 2002 is located at the R_V1?
If the labels corresponds to this LUT, editing the hcpmmp1_original.txt to these indexes (starting from 1001 instead of 1000 and similarly for 2001 onward) should give you the correct output.