[Insight-developers] itkAnalyze Valid Orientations and 3D - 2D conversions

[kent williams]

This bug: http://www.vtk.org/Bug/view.php?id=8650 isn't related to the
question Luiz brought up, it is about having the Analyze and NIfTI readers
handle images that (incorrectly) specify 0 as the number of dimensions in
the dime.dim field.

As for the actual questions:

On 4/2/09 12:07 PM, "Luis Ibanez" <luis.ibanez at kitware.com> wrote:
> 
> 
> When starting to test the 3D and 2D combinations, the following
> questions arise:
> 
> 
> 1) When instantiating a 2D image (that we plan to write out).
> 
>     What are the orientations that
>     Analyze is supposed to support ?
> 
I assume you've done some Googling for information; the most helpful links
are

http://www.grahamwideman.com/gw/brain/analyze/formatdoc.htm
http://eeg.sourceforge.net/ANALYZE75.pdf

Neither are particularly helpful when it comes to 2D.  As you wrote, only 3
orientations are supported in Analyze7.5 files; and the orientation field is
not well-defined anywhere for 2D.

You run into the same problem we've discussed elsewhere -- 2D Direction
Cosines don't contain sufficient information to make ANY statement about
their orientation. I'd go so far as to say 2D direction cosines are
completely useless in defining anything about orientation.

As this applies specifically to the Analyze 7.5 format, you could fill in
the hist.orient on write with your best guess, but who knows what other
applications will make of this.

Even if you look at the NIfTI format, you still have issues, because NIfTI
only has 3D Orientation in the header. At least it is well defined, but in
the case of 2D Images, it still expects a 3D orientation.  Or nD for n > 3,
still 3D orientation.

> 
> 2) When reading a 3D image in any of the orientations listed
>     above (RPI, RIP, PIR), we are assuming that the resulting
>     2D image will simply have an 2D orientation matrix with
>     an identity matrix. Given that there is no way to represent
>     its original 3D orientation by using a 2D matrix.
> 

That's what we ended up deciding to do as a complete punt.

I'd go so far as to say that there's no point in even WRITING an Analyze
file of a 2D image, unless you know the orientation assumptions of the
program you expect to read the image.  If you know those assumptions, you
can write the 2D analyze image in the proper raw pixel arrangement, and
figure the hist.orient field doesn't matter.



Notice: This UI Health Care e-mail (including attachments) is covered by the Electronic Communications Privacy Act, 18 U.S.C. 2510-2521, is confidential and may be legally privileged.  If you are not the intended recipient, you are hereby notified that any retention, dissemination, distribution, or copying of this communication is strictly prohibited.  Please reply to the sender that you have received the message in error, then delete it.  Thank you.