FA decrease after MNI transformation

Dear MRtrixers,

I am trying to calculate mean FA values of corticospinal tracts (CST) in glioma patients in MNI space. I noticed that my FA values always decrease after MNI transformation, with an average mean difference of -0.08 in MNI space compared to subject space (n=117). I think this difference is too large to simply be attributed to “noise” introduced during the tranformation to MNI. Does someone know how to explain / avoid this decrease in FA values in MNI space? It appears that larger tumors / tumor closer to the CST introduce a greater “FA loss” in MNI space compared to tracts that visually seem unaffected by for instance more distant tumors. Could the transformation be less accurate when the patient brain to fit onto the MNI brain is more disrupted by a tumor, thereby introducing these FA losses?

On visual inspection, the MNI transformed .tck files appear to be correctly transformed, and the MNI FA maps have a good overlap with the MNI T1 brain. I used the ANTs method to register patient T1 to MNI, after which .tck files where tranformed using the Scilpy tract transformation, with a deformation field provided.

DWI preprocessed scans were transformed to MNI using the Scilpy image transformation, after which I used the method described in this topic to calculate mean FA values.

We use MNI space since we want to include streamline coordinates in our analyses as well, and compare our patients to HCP data. Looking forward to your expert opinions!