FreeSurfer White Matter Parcellation¶

The -f flag in DCE_all.sh enables FreeSurfer-based white matter parcellation, which divides the WM mask into anatomical subregions. This allows Ktrans to be analyzed separately in specific WM tracts and lobar regions rather than as a whole-brain average.

Overview¶

Standard DCEPrep analysis computes Ktrans statistics for the whole WM and GM masks. With -f enabled, FreeSurfer's wmparc atlas is used to subdivide the brain into anatomically meaningful subregions, providing region-specific Ktrans and Vp statistics in the population report and spreadsheet export.

Requirements¶

• FreeSurfer must be installed (v6.0.0, Linux-centos6_x86_64-stable-pub-v6.0.0-2beb96c)
• FreeSurfer license.txt must be available (mounted in Docker)
• FreeSurfer recon-all must have been run for each subject (outputs in derivatives/freesurfer/)
• Sufficient disk space (~300 MB per subject for FreeSurfer outputs)

Enabling Parcellation¶

./DCE_all.sh -d /data/rawdata -f

Parcellation runs as part of the analysis pipeline's masking step.

Parcellation Atlas and Registration¶

DCEPrep uses FreeSurfer's wmparc (white matter parcellation) atlas, which combines cortical parcellation labels from the Desikan-Killiany atlas (aparc) with subcortical segmentation from aseg.

The registration pipeline:

1. mri_label2vol converts wmparc.mgz from FreeSurfer conformed space to the subject's native T1w space (rawavg)
2. mri_convert converts the result to NIfTI format
3. antsApplyTransforms applies the T1w → DCE rigid transform (with nearest-neighbor interpolation) to bring the parcellation into DCE space
4. The result is saved as anat/sub-##_ses-##_space-DCEref_desc-wmparc.nii.gz

Output Subregions¶

The following regions are extracted from the wmparc atlas and reported in the population report and spreadsheet:

Subcortical structures (from aseg)¶

White matter regions (from wmparc)¶

Cortical regions (from aparc)¶

Composite region¶

For each region, the population report includes median Ktrans, median Vp, and volume (mm³). The spreadsheet also includes cortical thickness (average and std) for all Desikan-Killiany parcellation regions.

Run Time¶

FreeSurfer surface reconstruction is computationally expensive. Expect ~6–8 hours per subject on a standard workstation. The -s (skip) flag can be used to avoid re-running parcellation for already-processed cases.