[MINC-users] Best plane (axial or coronal) for CIVET processing?

[Andrew Janke]

On 23 March 2012 22:48, Alex Zijdenbos <zijdenbos at gmail.com> wrote:
> I can take a stab at this - Andrew can correct me if I am wrong somewhere.

Seems pretty good to me.

> From the 'register' screenshot I would say that the image on the left
> (I assume that is the 'axial') is converted correctly (barring a
> left-right inversion that wouldn't be visible); and the second one is
> incorrectly converted.

Correct.

> I suspect that nii2mnc (or ana2mnc) has no way of knowing that you
> flipped the axes around

Correct.  I have seen instances of ANALYZE where files are stored in
the order they came of the scanner. Of course if your data starts off
being wrong. ie: you've lost the co-ordinate information when you
converted from DICOM? to UNC to ANALYZE7.5, no amount of automated
conversion fiddling
is going to save you.

> You can probably fix this by providing ni2mnc with one of
> the spatial dimension ordering options, such as -coronal, to tell it
> what the ordering of the dimensions in your image is.

Correct. But again, while this will result in a conversion that
"looks" good you still have no way of knowing if a left-right flip has
been done as part of the conversion to ANALYZE some time ago. There is
no way to solve this problem as you would have to know exactly which
path the original file took through SPM/ANALYZE/FSL/etc and know the
defaults that were set in each of these in order to figure it out.

> All MINC tools display data in neurological convention (left=left).

Again, correct. This I can't stress enough is only a viewing
convention. For an axial slice, radiological means you are viewing the
axial slice from the patients foot and neurological means you are
viewing from the top of the patients head. The way a radiological department
used to "swap" between the viewing conventions was to flip the X-ray over
on the viewing desk! The same principle applies
for coronal and sagittal. The data itself should never be flipped L-R
but unfortunately some other tools do this.

>> So, could you please tell me whether the nii2mnc conversion has been done
>> properly?

Well yes, the conversion from nii2mnc very likely was correct for both
images. The problem is the input ANALYZE or UNC file is poorly or more
likely not defined WRT co-ordinate systems.

>> Please be reminded that my interest is not in having both axial and
>> coronal MINC images, but just making sure the UNC-to-Analyze-to-MNC has been
>> done successfully whether it's is in axial or coronal plane.

>From what I have seen, this will not be possible with your data. You
will need to manually inspect each ANALYZE dataset and use the
appropriate conversion depending on if it is a "axial" or a "coronal"
ANALYZE file.

If you get stuck, and only as a last resort, use the attached scripts.
These are my swiss-army-chainsaw-botched-file-fixer-upperers, and should be
used with a great deal of caution! It looks to me as if your "coronal"
file needs to first have the Z and Y axes swapped and then be flipped
in Y. This is a fix I have seen a few times before:

   $ volmash -swap yz coronal.mnc tmp.mnc
   $ volflip -y tmp.mnc fixed.mnc


a