[Mrtrix-discussion] Spatial info and image orientation

[Donald Tournier]

Hi Alessandro,

The data scaling field refers to the offset and multiplier that are applied
to the data values as they are read from the file. This allows data stored
in integer formats to represent values outside the range they would
otherwise be able to represent. For example, you could use a 16 bit integer
(as is typical is DICOM) with a maximum value of ~65,000 to represent
values in the range 0 to 1 with the appropriate multiplication factor. The
DICOM standard allows for this, as does NIfTI & analyse.

As to your second question, this has to do with the many ways voxel
intensities can be stored in a file. The layout field  refers to the
relationship between where in the file a voxel intensity is stored (i.e.
its file offset) and where it is within the image volume. So it's perfectly
feasible for the axis to be reversed in the layout field without affecting
the final image - as long as the voxel intensities themselves have been
correspondingly reordered on file. MRtrix will always do this when
converting to NIfTI, so that the layout is 'standard' - i.e. as close to a
axial orientation as possible. Note that the transform field also plays a
part in this - if the dataset consists of sagittal slices, the transform
should incorporate the rotation required to keep the relationship
consistent. When converting to NIfTI, MRtrix will reformat that to a
near-axial orientation, make the layout +0,+1,+2, modify the transform to
match, and reorder the voxel intensity values accordingly. The important
thing is everything remains self-consistent...

Of course, if you flip the image using other means without reordering the
data appropriately, you will affect that relationship and compromise the
self-consistency of your data - something I really advise against. I have
spent far too much time trying to sort out ambiguities in the left-right
convention of the analyse format, so I've made sure MRtrix behaves
correctly. Even if the layout does change, your images should remain
unaffected, at least within MRtrix - I know many people have been surprised
to find images processed using MRtrix don't match up with the originals in
FSLView, for example. That's not because the images are incorrect, is
because FSLView assumes both images have been stored exactly the same way.
As you found out, MRView had no such problem.

If you need more details about these issues, I think there is a more
in-depth explanation somewhere in the MRtrix documentation, in the
'supported image formats' section, if I recall correctly.

Hope this helps.
Cheers,

Donald

--
Dr J-Donald Tournier (PhD)

Senior Lecturer, Biomedical Engineering
Division of Imaging Sciences & Biomedical Engineering
King's College London

A: Department of Perinatal Imaging & Health, 1st Floor South Wing, St
Thomas' Hospital, London. SE1 7EH
T: +44 (0)20 7188 7118 ext 53613
W:
http://www.kcl.ac.uk/medicine/research/divisions/imaging/departments/biomedengineering

On 24 Jan 2014 16:57, "Alessandro Calamuneri" <alecalamuneri at gmail.com>
wrote:

> Hi all,
> I have a question about image spatial orientation.
> First of all, when looking at mrinfo output, what does multiplier into
> data scaling mean?
> Here the main question.
> I have two images, an fa calculated using tensor2fa, and another image
> coming from the same dataset but obtained from a different package. When
> looking at spatial infos, I find transformation matrices are the same,
> whereas data layout is +0 +1 +2 for fa and -0 +1 +2 for the other one.
> Despite of that, when I overlap them into mrview they are not flipped.
> Should not they be flipped considering last info?Or does mrview consider
> something different for image displaying?
> On the other way around, if I try to use fslorient and fslswapdim to get
> the same orientation, I find that image properties become the same, whereas
> volumes are flipped into mrview.
> How is it possible?
>
> Best,
> Alessandro
>
>
> _______________________________________________
> Mrtrix-discussion mailing list
> Mrtrix-discussion at www.nitrc.org
> http://www.nitrc.org/mailman/listinfo/mrtrix-discussion
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.nitrc.org/pipermail/mrtrix-discussion/attachments/20140126/2af64f44/attachment.html>