Hi,
No - scaled "mm" always used the positive, unsigned pixdims.
Signed pixdims are horrible and a real hack in Analyze - thankfully
they are banned in nifti!
All the best,
Mark
On 1 Sep 2006, at 12:54, Ged Ridgway wrote:
> Dear Mark,
>
> Sorry, another question... the scaled coordinates -- do they use
> *signed* voxel dimensions? E.g.
> avwhd $FSLDIR/etc/standard/avg152T1.hdr
> reports the pixdims as (-2, 2, 2) so does this mean that the scaled
> coordinate of voxel (10,10,10) would be (-20,20,20), or would it be
> (20,20,20) using absolute pixdims?
>
> Times like these, I wish I worked with nice simple scalar time-series
> data... I hate coordinate system "standards"!
>
> Many thanks,
> Ged.
>
> Mark Jenkinson wrote:
>> Dear Ged,
>> Yes, our "mm" coordinates at the moment are scaled voxel coordinates.
>> The easiest way to do things at the moment is to use voxel coordinates
>> as it is less confusing.
>> As for the future - the plan is for the user to only ever deal with
>> the same "mm" coordinates that fslview reports. This would be the
>> nifti qform or sform world coordinates when these were set, and
>> otherwise it would default to our scaled coordinates, just as fslview
>> currently does in its reporting. However, there are quite a few
>> places in our code that need to be updated and tested for this, so
>> it didn't quite happen in the last release. Hopefully it will in the
>> next release.
>> I'll put the fslview request on to the appropriate list.
>> All the best,
>> Mark
>> On 1 Sep 2006, at 11:26, Ged Ridgway wrote:
>>> Hi,
>>>
>>> Am I right in thinking that if I use pointflirt to compute a
>>> transformation from corresponding "mm" coordinates (i.e. I don't use
>>> the --vox option and specify -i & -r volumes) these should be
>>> "scaled"
>>> coordinates (slide 127 of
>>> http://www.fmrib.ox.ac.uk/fslcourse/lectures/flirt_fugue/
>>> flirt_fugue_slides.pdf)
>>> and not "world/standard space" coordinates as reported from e.g.
>>> fslview?
>>>
>>> Is FLIRT/pointflirt likely to change in the future to using the
>>> NIfTI qform for the voxel-world mapping, so that saved xform.mat
>>> files map from world-to-world rather than scaled-to-scaled?
>>>
>>> Also, a feature request: it would be very useful to have an fslview
>>> button that appends the current cursor position (in scaled
>>> coordinates, if that's appropriate) in the current image, to a file,
>>> suitable for use with pointflirt. So, e.g. one could load up two
>>> images, find a point, click the button (enter a filename), switch
>>> images, find the corresponding point, click, repeat several times,
>>> then run pointflirt -- without having to manually multiply voxel
>>> coords by their voxel dims and enter them with a text editor.
>>>
>>> Best,
>>> Ged.
|