Hi Mark,
Sorry - for some reason I was thinking that RAS = radiological. That
explains a lot of my confusion. Now almost everything is consistent.
Are you sure FSLView only swaps left and right if the determinant is
negative? I see it swap L/R for positive qform determinants. I haven't tried
feeding it a negative one yet.
Jolinda
-----Original Message-----
From: FSL - FMRIB's Software Library [mailto:[log in to unmask]] On Behalf
Of Mark Jenkinson
Sent: Thursday, April 20, 2006 1:43 PM
To: [log in to unmask]
Subject: Re: [FSL] Yet another orientation question
Dear Jolinda,
Just to clarify:
We consider "neurological" to have positive determinant.
We consider "radiological" to have negative determinant.
We treat *all* Analyze files as being "radiological".
I believe that this is all consistent with the FAQ and what you see.
If you convert Analyze to Nifti and it is originally RAS ordered, then you
need to make sure that you set an sform/qform appropriately. The default
one provided by avwchfiletype is based on our assumptions of LAS. You can
easily change this by using avworient -swaporient.
Currently in FSLView we simply swap left and right if the determinant is
negative. This isn't ideal in many ways, but we don't yet have the
capability to do arbitrary transforms on the fly, so we've left it at that
(although doing arbitrary transforms based on the sform/qform is on our todo
list).
We have this behaviour so that it is possible to see what happens when the
sform/qform matrix information is changed. This is particularly important
for Analyze files and for nifti files where no sform or qform is set. In
this case we feel it is important to show how FSL will interpret the
left-right order (our default choices).
So although this isn't the behaviour you'd most like from FSLView (and
slices) I think it is behaving correctly. Does this accord with your view?
I hope this is clear.
All the best,
Mark
On 20 Apr 2006, at 19:38, jolinda wrote:
> Thanks Mark. I just want to make sure that my converted files are
> correctly
> interpreted by FSL. I haven't run them through all the tools yet,
> but I have
> found a few things.
>
> With older analyze files, running avwchfiletype NIFTI adds the
> following
> sform information:
>
> sform_name Aligned Anat
> sform_code 2
> sto_xyz:1 -3.125000 0.000000 0.000000 96.875000
> sto_xyz:2 0.000000 3.125000 0.000000 -96.875000
> sto_xyz:3 0.000000 0.000000 4.000002 -60.000031
> sto_xyz:4 0.000000 0.000000 0.000000 1.000000
> sform_xorient Right-to-Left
> sform_yorient Posterior-to-Anterior
> sform_zorient Inferior-to-Superior
>
> Because of this, if the data was originally RAS, it will now be
> left-right
> flipped in all of the analysis reports. Is this based on
> information in the
> analyze header, or an assumption about the order of the data? I
> understand
> that some assumptions must be made, but it appears to contradict the
> statement in the FSL faq that the latest version of FSL treats all
> Analyze
> images as radiological.
>
> FSLView appears to interpret the qform matrix in the same way that
> avworient
> -getorient does, and to flip LR accordingly. Thus, if you feed it:
>
> qform_name Scanner Anat
> qform_code 1
> qto_xyz:1 3.130000 0.000000 0.000000 0.000000
> qto_xyz:2 0.000000 3.130000 0.000000 0.000000
> qto_xyz:3 0.000000 0.000000 4.000000 0.000000
> qto_xyz:4 0.000000 0.000000 0.000000 1.000000
> qform_xorient Left-to-Right
> qform_yorient Posterior-to-Anterior
> qform_zorient Inferior-to-Superior
>
> or:
>
> qform_name Scanner Anat
> qform_code 1
> qto_xyz:1 -3.124945 -0.001332 -0.023744 101.946167
> qto_xyz:2 0.009799 -2.767111 -1.858704 103.824890
> qto_xyz:3 -0.015807 -1.452144 3.541846 2.113121
> qto_xyz:4 0.000000 0.000000 0.000000 1.000000
> qform_xorient Right-to-Left
> qform_yorient Anterior-to-Posterior
> qform_zorient Inferior-to-Superior
>
> using the same raw data, it will display the image with left and right
> flipped (when compared with the same data with qform_code = 0).
>
> I do not think this is good behavior. I think it should either
> ignore the
> qform matrix altogether, or use it to transform the image into RAS
> space.
> Right now it seems to simply look at the determinant of the matrix and
> decide whether to flip left and right -- and decide wrongly.
>
>
> Jolinda
>
> -----Original Message-----
> From: FSL - FMRIB's Software Library [mailto:[log in to unmask]] On
> Behalf
> Of Mark Jenkinson
> Sent: Thursday, April 20, 2006 9:50 AM
> To: [log in to unmask]
> Subject: Re: [FSL] Yet another orientation question
>
>
> Dear Jolinda,
>
> This issue is confusing and our documentation on this tries to walk
> the line
> between being comprehensive and comprehendable by the majority of
> users. I
> think that the issue of left-handed and right-handed coordinate
> systems is
> somewhere between incomplete, confusing and wrong. I think I will
> remove
> these statements as they are not helpful to most people, and the
> people who
> do appreciate thinking in these terms want more information than we
> provide.
>
> You are right about the nifti xyz coordinate system being RAS and
> therefore
> right-handed. It is also true that LPS is right-handed.
>
> What is the crucial part of our "radiological/neurological" definition
> is the
> mapping between two coordinate systems: the voxel coordinates (not
> in real
> space) and the mm (real space) coordinates. This *mapping* is what is
> important for us. Therefore we define radiological/neurological
> (terms
> which aren't very good here, but are used for familiarity and lack
> of better
> terms) on the basis of the sign of the determinant of the matrix
> that maps
> between these two coordinate systems. So I always think in these
> terms and
> not in terms of the handedness of the coordinate system (as there
> are too
> many potential coordinate systems).
>
> Sorry that this has been a source of confusion for you.
> I hope that this clears things up.
> I will edit the FAQ to remove the handedness text shortly.
>
> All the best,
> Mark
>
>
>
>
>
> jolinda wrote:
>
>> Okay...but I'm afraid I just don't see why a positive determinant
>> indicates a left-handed coordinate system. The qform matrix
>> relates the
>> voxel coordinates to the NiFTI standard right-handed RAS space ---
>> so,
>> if the image was already in the same space, the qform would be
>>
>> 1 0 0 0
>> 0 1 0 0
>> 0 0 1 0
>>
>> which certainly has a positive determinant. If you are using a qform
>> code = 1, the you are most likely relating the standard DICOM space
>> (LPS) to NiFTI space (RAS), both of which are right-handed, and your
>> qform matrix will indicate a 180 degree rotation and still have a
>> positive determinant.
>>
>> Can you explain this to me? What am I missing?
>>
>> Jolinda
>>
>>
>> -----Original Message-----
>> From: FSL - FMRIB's Software Library [mailto:[log in to unmask]] On
>> Behalf Of Christian Beckmann
>> Sent: Wednesday, April 19, 2006 1:05 PM
>> To: [log in to unmask]
>> Subject: Re: [FSL] Yet another orientation question
>>
>>
>> Hi,
>>
>> it's the qform matrix which counts, see http://www.fmrib.ox.ac.uk/
>> fslfaq/#general_radiologicaldef
>>
>> "Our technical definition of neurological convention for image
>> storage is that the coordinate system is left-handed, or
>> alternatively, that the determinant of the sform or qform matrix that
>> relates real world coordinates to voxel coordinates, has a positive
>> determinant."
>>
>> In your case, the qform matrix has a positive determinant and the
>> volume is therefore understood to be in neurological convention.
>> cheers
>> christian
>>
>>
>> --
>> Christian F. Beckmann
>> Oxford University Centre for Functional
>> Magnetic Resonance Imaging of the Brain,
>> John Radcliffe Hospital, Headington, Oxford OX3 9DU, UK
>> Email: [log in to unmask] - http://www.fmrib.ox.ac.uk/
>> ~beckmann/
>> Phone: +44(0)1865 222551 Fax: +44(0)1865 222717
>>
>>
|