We are calculating MTR twice on a single subject in the same session. We have two sets of MT/noMT protocols (MT1,noMT1,MT2,noMT2) that all need to be aligned for the final MTR diff. Our current strategy is to align all four to the structural and then do the voxel calculations.
But, after flirt, the images are not aligned.
This can be seen best in an animation of `slicer -a` for both struct aligned MT images: http://imgur.com/a/KxD7s .
all images are here: http://imgur.com/a/TgS0S
Is this the best we can expect?
Are there other techniques to tighten the alignment?
Can we approach the 4 way alignment in a better way?
We've tried with different costs (corratio vs normmi) and dof (6 vs 7) values. e.g.
flirt -cost normmi -in MT1_bet_orient -ref struct1_bet_orient -out MT1_struct1_normmni_dof7_spline -dof 7 -interp spline
flirt -cost normmi -in MT2_bet_orient -ref struct1_bet_orient -out MT2_struct1_normmni_dof7_spline -dof 7 -interp spline
We also tried mcflirt on the concatenated the MTs without any improvement. Our next stab at it is to provide a cost function weighting favoring white or grey matter voxels (manipulated structural fast output).
the pixdim1-4 for the files are:
MT1_bet 1.796875 1.796875 1.800000 1.000000
MT2_bet 1.796875 1.796875 1.800000 1.000000
struct1_bet 1.000000 1.000000 1.000000 1.000000
MT1_struct1_normmi 1.000000 1.000000 1.000000 0.000000
MT2_struct1_normmi 1.000000 1.000000 1.000000 0.000000
Thanks!
