Dear George, The volumes in the text files come from soft segmentations at high resolution. There are two factors contributing to the discrepancies:

1. The downsampling to 1mm isotropic 2. The use of soft vs hard segmentations when computing volumes. For example: if you’ve got 100 voxels with 1% probability of being in the VM, the volume will be one voxel. But if you count the number of voxels being labeled as VM, you’ll get zero. Cheers, /Eugenio

Juan Eugenio Iglesias Senior research fellow CMIC (UCL), MGH (HMS) and CSAIL (MIT) http://www.jeiglesias.com

From: freesurfer-bounces@nmr.mgh.harvard.edu on behalf of "Thomas, George" george.thomas.14@ucl.ac.uk Reply-To: Freesurfer support list freesurfer@nmr.mgh.harvard.edu Date: Thursday, April 15, 2021 at 07:07 To: "freesurfer@nmr.mgh.harvard.edu" freesurfer@nmr.mgh.harvard.edu Subject: [Freesurfer] Thalamic segmentation in FS voxel space

External Email - Use Caution

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