Dear FreeSurfer users,

I'm using FS 6.0. I registered DWI scans to the CVS registrations of T1-weighted MRI scans as described at https://surfer.nmr.mgh.harvard.edu/fswiki/FsTutorial/DiffusionV6.0. The average diffusion coefficient (ADC) maps registered to CVS have a "halo" of very low values around the brain (right outside the pial surface of the cortex), which diverges from the expectation that ADC be lowest in white matter.

See attached image of ADC in CVS space for one sample: adc-ANAT+CVS-to-avg35.png

On the other hand, this halo of low values is not as apparent in the ADC maps in diffusion space (masked to voxels in subject's aseg in diffusion space):

See attached image of ADC in diffusion space: diffusionSpace-adc.png

For comparison, also attached are the subject's T1-weighted images registered to CVS space (final_CVSmorphed_tocvs_avg35_norm.png) and to diffusion space (diffusionSpace-brainmask.png) that underlay these spatial maps.

On the other hand, fractional anisotropy (FA) maps in either CVS space or diffusion space look reasonable --- with FA highest along white matter fiber tracts. See attached screenshots for FA in CVS space (fa-ANAT+CVS-to-avg35.png) and FA in Diffusion space (fa-diffusionSpace.png; masked to aseg in diffusion space).

This phenomena is typical across all samples. Is the halo of very low values surrounding the brain, adjacent to cortical gray matter, typical in ADC spatial maps registered to CVS?

If so, is there a way to remove this halo of low values for ADC maps in CVS space? For example, the depth of this halo decreases when masking the ADC maps with the aseg.mgz for the CVS average template. The halo decreases furthermore when masking by the subject-specific aseg registered to CVS generated by mri_cvs_register ("final_CVSmorphed_tocvs_avg35_aseg.mgz"). In either case, the halo does not disappear.

I'm concerned that these ADC mis-measurement e boundary may yield significant issues in GLM analysis. If there is no way to remove this halo, is a GLM of ADC maps in CVS space more reliable for voxels masked to white matter?

Thank you.

Best, Chintan