[MINC-development] minc 2.0 file format definition

[romain]

Hello

thank for you short response.

Robert VINCENT wrote:

>Romain,
>
>You are correct about equation (4).  I had updated it in the source TeX
>files but I had not updated the document on the website.  I will update
>the website to reflect this correction.
>  
>
(a pdf version of the website will be also great)

>Regarding the definition of the coordinate spaces, the meaning of the
>"true" xspace, yspace, and zspace dimensions is invariant with respect to
>the acquisition order of the image.  X is always from patient left to
>patient right, Y is from patient posterior to anterior, and Z is from
>patient to inferior to superior.  A MINC file may have any of these
>dimensions chosen as the fastest-varying dimension, so a coronal image and
>a transverse image may share precisely the same definitions of X, Y, and Z
>even though the organization of the data in the file may change.
>  
>
that is what I understand when I first read the minc format, and this is
also a reason to prefer the minc format to other one (dicom, nifti).
But I still do not understand how it works.

>The direction cosines simply represent (in a very general fashion) the
>rotation that must be applied to the voxel (IJK) coordinates in order to
>derive the world (XYZ)  coordinates.
>  
>
Yes but in this relation (which is the same as the one use for dicom
or nifti) the I,J,K is related to the order of varying dimension
(I beeing the fastest-varying dimension, then J then K).

So if a transverse and a coronal image share the same definitions of
Xspace, Yspace and Zspace, then they have the same transformation
matrix (the one define by equation (4) ).
Then the first point will have the same spatial localisation which is 
false !
(if the data order change the transformation matrix should also change ! ?)

How can you handle the same transformation matrix with different order 
organisation
of the data ?

Thank for your help

chears

Romain