dHCP T1w <-> T2w mis-corregistrations issue


first thanks for sharing this incredible dataset, the data quality is quite amazing !

I am working on automatic segmentation of T1w and T2w volumes, and I notice co-registration issues between the two volume (the mismatch is very subtle but enough to quantitatively impact gray matter (GM) comparison (when the GM is computed from T1 or T2 volumes)

So I wonder if there are some physical reason for different spatial deformation between the two acquisitions ?

Do you know if the sub-xxx_T2w.nii.gz and T1w.nii.gz have been corrected for spatial deformation (due to the scanner gradient non linearity) … I understood there was a special step for it for the young adult HCP dataset .

I guess there was also a step for coregistration of T1w toward the T2w volumes ? which tool was used ? (rigid affine on Non Linear)

Finally any hints to improve this coregistration ? (I already tried a few tool elastix from simple elastix and reg_aladin from niftireg) but I do not achieve perfect registration …

Many thanks for the help