[Insight-users] VectorImage Orientation from NRRD

[Gordon Kindlmann]

Hello,

Its not deference to authors that prevents changing the RAS->LPS  
orientation transition behavior you ran into- its backwards  
compatibility.  I agree that it might be better if readers were only  
that- readers- that did no other meta-information manipulation.

Perhaps there can be some discussion here about: (1) the extent to  
which LPS is really the presumed space of orientation representation  
for 3D images, and (2) suggested mechanisms for disabling the RAS- 
 >LPS orientation transformation.

For example, are there currently any other readers for complicated  
formats, where the caller can set some state in the reader to control  
what meta-information is returned by the reader, or how the meta- 
information is modified?

Gordon

On Oct 16, 2008, at 9:42 PM, Daniel Betz wrote:

> Hello ITK developers,
>
> An itk::Image that behaves the same like itk::OrientedImage will
> facilitate a very clean and nice interface to NRRD files, I am very
> glad about that.
>
> One thing remains. I am afraid to peeve too much with that and
> don't want to be misunderstood. I starred a long time on some
> coordinates on my screen and nothing made sence. And there are
> still questions.
>
> itkNrrdImageIO.cxx. I am using at the moment revision 1.30. The
> cvs commit entry for revision 1.31 was: "now handles LPS to/from
> coversion for 3D images".
>
> There is apparently no public information available about a
> convention for ITK to load orientated images into an unorientated
> image only in LPS space.
>
> Somebody told me, that people are "freaking out" when they get
> images with another orientation as DICOM. The sign flipping
> methode in the file loaders is definitly the wrong way to adress
> this issue. If the matter is, that people are not aware of the
> different coordinates, the application could make a red border
> around the window, tinker the coordinate letters or prevent
> loading or something like that if the world space is not LPS.
> Anyway, if conversion is "well defined" the application can do
> that.
>
> I remembered that Slicer3 was loading and showing my NRRD file
> properly and took a look into the file Slicer3/Libs/vtkTeem/
> vtkNRRDReader.cxx. I found a similar mechanism, probably from
> the same author; this time LAS and LPS are flipped to RAS.
> (by the way, the default branch of the switch statement is very
> enlightening). I don't know enouth about VTK to say if such a
> conversion makes sence here, but I am in doubt.
>
> My impression is, that the transformations in the file readers
> were a mistake, which nobody wants to correct in deference to the
> author or something like that.
>
> Finally, I only want some insight, and I want it stable, reliable
> and especially predictable.
>
> Regards
> Daniel Betz