FIX: registration issue between T1w and BOLD in specific datasets

[HippocampusGirl]

Changes proposed in this pull request

• Switch to using mri_coreg in fsl_bbr_wf instead of flirt as suggested by @effigies and @oesteban in Failed registrations between BOLD and T1w #2591
• Add -basescale 1 parameter to the flirt boundary-based registration call to avoid the scaling behavior there too. As far as I can tell, this does not have side effects, but it may make sense to wait for a response from the flirt developers before merging this one.

Documentation that should be reviewed

No changes to documentation

[oesteban]

How does this work with your data @HippocampusGirl ?

[effigies]

Hmm. If we're adding -basescale 1 to the BBR step, then it feels like adding -basescale 1 to the initial FLIRT step is less intrusive than switching to mri_coreg.

Looking at the source:

void set_basescale(const string& filenameA, const string& filenameB)
{
  if (!(gOptions->force_basescale)) { 
    // only try to automatically determine if it was not requested by the user
    volume<float> volA, volB;
    read_volume_hdr_only(volA,filenameA);
    read_volume_hdr_only(volB,filenameB);
    float maxdimA = Max(Max(volA.xdim(),volA.ydim()),volA.zdim());
    float maxdimB = Max(Max(volB.xdim(),volB.ydim()),volB.zdim());
    if (Max(maxdimA,maxdimB)>12) {  // over 150% of largest (8mm) scale
      gOptions->basescale = Max(maxdimA,maxdimB)/8;
    }
    float mindimA = Min(Min(volA.xdim(),volA.ydim()),volA.zdim());
    float mindimB = Min(Min(volB.xdim(),volB.ydim()),volB.zdim());
    if (Min(mindimA,mindimB)<0.75) { // between 0.5 and 1 mm 
      gOptions->basescale = Min(mindimA,mindimB);
    }
  }
}

-basescale 1 only has an effect if a voxel dimension is > 12mm or <0.75mm, so I think we can be reasonably confident that the number of datasets affected will be quite small.

I'm okay moving to mri_coreg if everyone wants to keep on that way, but I think I have a slight preference for just setting -basescale 1 on both steps of the FSL pipeline for now, and then moving to a common pipeline in future versions.

Sorry for jumping one way and then the other.

[HippocampusGirl]

@oesteban
I have only tested this with single subjects so far, where the results are very promising. I will be re-running some larger datasets that had the issue over the next few days, and will get back to you about that.

@effigies
I am happy either way, but I would prefer the mri_coreg option, since it would make the workflows more consistent between running with and without --fs-no-reconall.

[oesteban]

@HippocampusGirl thanks much for the update and the effort.

Since setting -basescale 1 on both FLIRT processes seems trivial, why don't we change the base of this PR to master and add the parameter to maint/20.2.x as @effigies suggests? That's probably going to be the best call to minimize the workload of the LTS maintainers.

Would that be okay?

[mgxd]

can this be closed now that #2624 and #2625 are in?