Hi John, everyone,
I'm using spm5 to segment a NiFTI image which has:
- qform_code/name 2/aligned anat
- qform containing just voxdims and origin
- pixdim fields matching diagonal elements of qform
- sform_code/name 4/MNI_152
- sform appropriate 12dof matrix following affine registration to
avg152 using FLIRT (result appears fine using spm_check_registration)
at the prompt I see:
Running "Segment"
Warning: QFORM0 representation has been rounded.
> In @nifti/private/encode_qform0 at 27
In nifti.subsasgn>fun at 109
In nifti.subsasgn at 20
In spm_create_vol>create_vol at 80
In spm_create_vol at 16
In spm_preproc_write>preproc_apply at 84
In spm_preproc_write at 24
In spm_config_preproc>execute at 478
In spm_jobman>run_struct1 at 1384
In spm_jobman>run_struct1 at 1392
In spm_jobman>run_struct1 at 1392
In spm_jobman>run_struct at 1351
--------------------------
Done.
But I don't wish anything to be done to the qform or sform! I have
switched off affine registration in the segment options and am
outputting only the native GM, so expect the c1 image to simply have
the same NiFTI geometry headers as the original.
Checking the output with FSL's avwhd, I see that the sformcode has
been changed to 2-aligned anat, the sform itself has remained
unchanged, but the qform has been changed, and is not equal to the
sform either, and worse still, the pixdims have now been changed.
Any idea what's going on?? It seems to me that the qform and pixdims
should be left alone, and that the sform should only change if spm is
used to do affine registration, in which case the sformcode should be
changed to (or left as) MNI, not changed to aligned-anat. Perhaps this
is some weird back compatibility thing, but it seems most odd... I'm
worried that the altered pixdims in particular may screw up further
processing.
Thanks in advance for any help from anyone, best regards,
Ged.
|