[MINC-users] nii2mnc problem

[Andrew Janke]

>    Thanks Andrew.  I placed a couple of samples into
>  "/data/scratch/scratch1/jim/nifti".  There you will find volumes for 2
>  subjects: (1) subject 0002, who is reflective of most of my subjects,
>  and (2) subject 0008, who, following conversion using the same input
>  params as used for S0002, is upside-down (according to Display).  I
>  used the following to convert:
>
>  nii2mnc -xyz -flipz -flipx 0002-M-NC.nii

Hi Jim,

Well after a bit of head scratching I have an answer for you.  In
short, nii2mnc is doing the correct thing but for some reason, most of
these volumes appear to be using various strange combinations of -ve
direction cosines and then negative steps on top of this.  This is
where I would say the mapping between the mnc voxel to world transform
and the Nifti Q-form xfm (registration) sort of breaks down.

You will find that if you look at your two output mnc volumes in
register that the third column (the aggregation of the two) is in fact
in the correct orientation. Register always views data from the voxel
perspective and just updates the world co-ordinate information to
suit. If you turn "synch" on you will see what I mean.

The only real "fix" for this is to mincresample them to something
sensible like 100, 010, 001 as opposed to the current -100, 010, 00-1
wierdness.


a