Hi Andrew,
Since cleanrun1.nii is oblique, let's ignore it for now and
focus on qto_xyz for cleananat.nii.
The matrix does display a problem with Philips NIfTI datasets
that I have seen before. Consider the cleananat.nii qto_xyz
matrix that you posted:
0.0 0.0 1.0 -79.5
-0.429688 0.0 -0.0 -123.349518
0.0 0.429688 -0.0 74.199028
0.0 0.0 0.0 1.0
The basic thing I look at is whether an applied transformation
matrix will end up covering the coordinates 0,0,0 somewhere in
the middle of the dataset. These do not.
The first row looks good. Starting at coordinate -79.5, it
moves by 1.0 mm with each voxel step. So that starts at a large
negative coordinate and heads in the positive direction, towards
zero. Good.
The second row does not look good. Starting at -123.349518 it
then moves by -0.429688 mm with each voxel step, going even
farther in the negative direction. The third row starts at 74.2
and moves farther in the positive direction. Both are bad.
At any rate, you should contact Philips about this.
- rick