External Email - Use Caution
Dear All,
I have run segmentThalamicNuclei.sh on a set of subjects with 1mm isotropic T1 images.
I need to convert these labels into volumetric space for further analysis so have been using the output with the FSvoxelSpace suffix and converting .mgz to .nii using mri_convert
However, for some of the subnuclei there seems to be a large discrepancy between the surface-based volumetric measurement completed by quantifyThalamicNuclei.sh, and the number of voxels in the FSvoxelSpace output.
Across the whole dataset of ~130 subjects, the left and right VM nuclei have a mean volume of about 25mm^3 according to quantifyThalamicNuclei.sh but are only represented by about 2-3 voxels per subject in the FSvoxelSpace output and are sometimes absent entirely.
I have found this to be similar for the VAmc (~30mm^3) and CL nuclei (~40mm^3), both of which are represented many fewer voxels than one might expect (~6 and ~20 voxels, respectively).
Of course, it will not always be possible to represent a given surface in voxel-space and you would expect some differences but am wondering why the discrepancy might be so large in these cases, and if there is anything that might be done to remedy this?
Many thanks, George Thomas