Native DWI space or resample to T1 before ROI analysis?

Hi Experts!

I am trying to run some analyses on DTI data but I am not sure about the space in which to work with. Some studies do all the analyses in the native DTI space, but others resample the DTI to the T1, and extract mean values from there (atlas already in T1 space). I am quite confused about this. My DWI data is 2mm isotropic.

Are there any recommendations for either approach?

  • I have been using ANTS to do the rigid body registration between the b0 and T1.
  • Non linear registration between T1 and atlas using ANTS
  • Atlas is then transformed into native T1 space and then the to native b0 space.
  • Finally, I use fslstats to extract data.

Thank you!


When you have DWI aligned to the T1 image, you have the benefit of using Anatomically Constrained Tractography (ACT). QSIPrep, for example, aligns T1 to ACPC axis, and then aligns DTI to that image. This becomes useful for tractography with QSIRecon. If you do a linear rigid transformation to get transform DTI to T1 space you should be fine since you won’t be messing with native geometry. But without knowing what analyses you have in mind, it is hard to know what is best.