I think I'm beginning to understand the function of the qform matrix...
To calculate the flow field, it's handy to have rc1 images with
isotropic 1.5x1.5x1.5 voxels, (and the template needs to be the same).
The result is the flow field, the u_rc1_Template.nii image. But... in
the rc1 image, there is a qform matrix in .mat0, and in the u_rc1
image, there is also a qform matrix in .mat0. My guess is that they
cancel out in the calculation of rc1 --> Template. (because the mat0's
are *not* different in Template_N.nii or in c1*.nii).
To apply the flow field (to other images), the function of the qform
matrix would be then to change the flow field in such a way as to
allow it to be applied to raw images (i.e., with other voxel
dimensions and orientation than the u_rc*.nii and the Template_N.nii),
without having to resample the raw images first?
Thanks for all the help,
Regards,
Johan
On Fri, Dec 11, 2009 at 3:38 PM, John Ashburner <[log in to unmask]> wrote:
> There is a second matrix in the headers that is used to encode the
> relationship between the imported and original scans. If you do a
> check-reg between an original and an imported, then you don't see them
> in register - whereas if you do the same with a couple of imported
> scans, then you would see them in approximate alignment. This is
> because most of the SPM routines use the "mform" information in the
> headers, whereas the DARTEL code also uses the "qform" information,
> which relates the original and native images with each other.
>
> To answer your question: Yes. DARTEL would not apply the combination of
> flow fields and the affine matrices in the way that you'd hoped.
>
> With hindsight, I would probably have used the matrices the other way
> around - but it is slightly too late now. At the time, it made writing
> the code much easier.
>
> Best regards,
> -John
>
> On Fri, 2009-12-11 at 12:13 +0000, Dr. Maximilian Wagner wrote:
>> Dear John,
>>
>> thank you for the suggestion. Just to double check: If I understand this
>> correctly, DARTEL applies the second matrix (used for the alignment of the
>> scans during the import step) to allow for the application of the flow field
>> (e.g. estimated on T1 scans) to non-imported scans (e.g. DTI, fMRI) in the
>> native space for highdimensional normalization, under the assumption that
>> the original T1 scans and DTI/fMRI scans were collected in the same subject
>> specific native space.
>> Thus, if we apply the DARTEL flow fields with the "Dartel-Normalize-to-MNI"
>> tool (without specifying template_6.nii) to non-imported DTI scans that were
>> coregistered with imported GM segments (rc1 scans) rather than the "native"
>> c1 scans, DARTEL would wrongly apply the combination of flow fields and the
>> affine matrix to the DTI scans that are actually in the "inter-subject"
>> rigidly aligned native space rather than the individual native space?
>>
>> Best regards,
>> Max
>>
>>
>>
>>
>
>
> --
> John Ashburner <[log in to unmask]>
>
|