[Mrtrix-discussion] Transformations on tracks2prob results

[Donald Tournier]

Hi René,

Yes, the transform from image grid to real (scanner) space is stored in the
NIfTI header. There are utilities to read those structures (there is a
nifti-bin package on Ubuntu that looks like it might do the trick).

Alternatively, you can read the transform using the mrinfo command from
MRtrix, although MRtrix will 'massage' the transform somewhat. Basically, it
scales each column of the rotation part by the reciprocal of the
corresponding voxel size, so that it maps from image space (i.e. in mm
relative to the image axes) to real space. Also, if the images are not
axial, the transform will be re-organised by switching axes to match the
equivalent axial. Hope that makes some kind of sense...

Cheers,

Donald.


2010/10/29 René Besseling <r.m.h.besseling at gmail.com>

> Dear all,
>
> I think Donald is right with the relative coordinates; the "mat" structure
> related to a 128x128x60 NIFTI with 2mm voxels is [-2 0 0 129; 0 2 0 -129; 0
> 0 2 -61; 0 0 0 1] and considering the last column this must be a relative
> transform from grid to coordinates a not an absolute one.
>
> Anyone any idea whether the absolute coordinates (with respect to the
> magnet isocentre or so) are stored in NIFTIs?
>
> Best regards,
>
> René
>
>
> 2010/10/27 Donald Tournier <d.tournier at brain.org.au>
>
> Hi René,
>>
>> In my (admittedly limited) experience with FLIRT, what *I think* I've
>> figured out is that the transform matrix maps from image space in A to image
>> space in B, both expressed in mm relative to their own image axes, not
>> relative to real (scanner) coordinates. I don't expect it to handle a shift
>> in the origin of either datasets, or for that matter any differences in
>> their respective transforms. That said, I'm not 100% sure I fully understand
>> all this, and I could easily be way off the mark... Maybe others can help
>> out who have more experience with FSL?
>>
>> Cheers,
>>
>> Donald.
>>
>>
>>
>> 2010/10/27 René Besseling <r.m.h.besseling at gmail.com>
>>
>>>   Dear MRtrix users,
>>>
>>>
>>>
>>> I’ve experienced some problems in transforming tracks2prob results using
>>> FLSs FLIRT.
>>>
>>> What I try to do is to compare the tractography result for two different
>>> scan sessions a and b, e.g. for the corpus callosum (cc). First I define a
>>> seed region in the cc based on the FA map of session a. Using this seed
>>> region I calculate my tracks. Next, I calculate the transform from FAa to
>>> FAb using FSL’s FLIRT. Let’s call this transform FAa2FAb. I use this
>>> transform to convert my session a seed to the session b seed and calculate
>>> my tracks. I convert both tractography results to tract density maps using
>>> tracks2prob. So far so good.
>>>
>>> However, in order to check the interscan agreement, I want to transform
>>> my session a track density map to my session b space. I do so by using the
>>> transform FAa2FAb. If I do so, my transformed tract density map is displaced
>>> over several cm with respect to where you would expect it to be looking at
>>> FAb. This is strange since FAb and FAa2FAb nicely agree (proper transform.
>>>
>>> I guess this has something to do with the origin of the tract density
>>> maps. Since I want to calculate my tract density maps at a different
>>> (higher) resolution than my FA maps, I don’t use the FA maps as reference
>>> but call tracks2prob with the argument –vox .5. This means the tract density
>>> images are calculated at a 0.5mm grid and tracks2prob automatically
>>> determined what size this grid should be based on the track coordinates.
>>> Because in both sessions different tracks are reconstructed, these grid
>>> sizes will differ between the 2 session. This should not be a problem since
>>> the 4x4 FLIRT transform works on coordinates, not on grids. However, as
>>> said, there seems to be a translational mismatch.
>>>
>>>
>>>
>>> Any ideas on how to solve this? Note that the transformation of my seed
>>> from session a to session b is OK, probably because seeda has the same
>>> resolution and size as FAa.
>>>
>>>
>>>
>>> Best regards,
>>>
>>>
>>>
>>> René Besseling
>>>
>>> _______________________________________________
>>> Mrtrix-discussion mailing list
>>> Mrtrix-discussion at www.nitrc.org
>>> http://www.nitrc.org/mailman/listinfo/mrtrix-discussion
>>>
>>>
>>
>>
>> --
>> Jacques-Donald Tournier (PhD)
>> Brain Research Institute, Melbourne, Australia
>> Tel: +61 (0)3 9496 4078
>>
>
>


-- 
Jacques-Donald Tournier (PhD)
Brain Research Institute, Melbourne, Australia
Tel: +61 (0)3 9496 4078
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.nitrc.org/pipermail/mrtrix-discussion/attachments/20101029/713c82fc/attachment.html