Map multiple modalities to a template space

We usually map through the high-resolution T1, per-modality brain extraction and "the right" modality for each type of image. I use antsr for all this and the procedure is roughly:

• call getMask on both modalities (or usually just the non-T1 modality - for T1, i use the ants cortical thickness results )

• employ the SyNBold parameters in the antsr antsRegistration program

• fixed image is the target modality - this saves time and space.

  • for dwi, i usually use the B0 image, possibly negated, possibly an average of B0 depending on data
  • for bold, i usually use mean bold
  • same as bold for ASL
  • for other modalities, modify brain extraction or registration approach as needed

• finally, save transformations and compose all maps to the template and transform / check results. use antsApplyTransforms for this.

one advantage of this approach is that label in template space can be transferred mostly consistently across modalities - but this assumes small/minimal error in the intra-subject+inter-subject mappings. the degree to which this is achievable will always depend on data quality.

Examples of this include the PTBP and the Neurobattery. These strategies evolve a bit with each dataset but are largely the same as what's been done for years.