Project DTI metrics onto T1 freesurfer recon-all middle surface of pial and white surfaces


I am thinking about projecting DTI metrics onto FreeSurfer cortical mid surface, the reason why the mid surface is to avoid the partial volume fraction effect.

After running recon-all, we got the vertex-corresponding thickness file, like rh.thickness.fwhm10.fsaverage.mgh, so basically, I want to do the same thing, but for DTI metrics, so that I can do some vertex-to-vertex group analysis.

Any idea would be appreciated


This is very strightforward using Freesurfer tools:

  1. Compute a registration matrix between you T1 and DTI (I recommend the B0). You can use both bbregister or mri_coreg.

  2. Use mri_vol2surf to project DTI metrics to cortex using the previously computed registration matric. At this step, you can choose which portion of the cortex to use. For example:
    mri_vol2surf --src /path/to/FA.nii.gz --out FA_in_cortex.nativespace.mgh --srcreg /path/to/regmatrix.lta --hemi lh --projfrac 0.5

  3. Move to standard space with mri_surf2surf

However, this might be a “not sufficient” approach to get rid of partial volume effects, since sometime voxel resolution can be wider than cortical ribbon. I would suggest to take a loot to this two publications to get rid of partial volume effects:

[ 1 ] Pasternak O; et al. Free Water elimination and mapping from diffusion MRI. doi:10.1002/mrm.22055

[ 2 ] Koo BB; et al. A framework to analyze partial volume effect on gray matter mean diffusivity measurements. doi:10.1016/j.neuroimage.2008.07.064

Also, FreeSurfer provides a “pipeline” to compute partial volume effects in PET, that might be useful (here),

Hi, @slieped
thanks very much, that is what I really want. Actually, I tried the petpvc in FreeSurfer with pet image, but I wonder if that works for dMRI,

Besides the cortical vertex-wise analysis, I also did subcortical and cortical ROI analysis based on FreeSurfer Desikan atlas.
Actually, I have already registered my FA map onto FreeSurfer conformed space successfully, but the statistical analysis between two groups (Control and Patients) gives me nothing for this approach based on the subcortical ROI and cortical ROI based on FreeSurfer desikan atlas.

Alternatively, the anatomical analysis of T1 did give me differences ROIs between the two groups.

So my question is that for neurodegenerative disease, if we found some macroscopic change between CN and PT in T1 image, normally we should also found something with DTI metrics which focuses on microstructure abnormalities, (microstructure damage should emerge before structural change) I think the results that I got for FA in grey matter microstructure change is maybe because of the partial volume fraction and CSF contamination like you said, I will read the two papers that you gave me.

Also, any suggestion would be appreciated:slight_smile:

I would focus on MD instead of FA, due to microstructure organization in cortial areas.

It is true that cortical ROI analysis would mitigate possible outliers (very sensitive in cortical approaches since CSF have muuuch higher MD values than GM and WM, but you might also lose spatial resolution.

In ROI analysis you could use approaches such as the ones described here to mitigate posible partial volume effects / outliers caused by mis-registration errors.
I also suggest to take a look to this new paper where they propose a non-linear warping optimized for cortical alignment using EPI images.

Victor M