users > image does not have orientation meta informat
Showing 1-3 of 3 posts
Results per page:
Feb 2, 2012  06:02 PM | Torsten Rohlfing
image does not have orientation meta informat
Regarding your question - here's a bit of background.

For historic reasons, CMTK orients all images into "RAS" space
internally. That means, the pixel array is kept in memory such that the
fastest-varying index is (approximately) along the anatomical left/right
direction, the second fastest index is along the posterior/anterior
direction, and the slowest is inferior/superior.

Clearly, these conventions only apply to human data (and to some extent to animal data).

Also, while the necessary anatomical reference information is usually
present in DICOM files and can be preserved when converting these to
NIFTI, NRRD, or (to lesser extent) Analyze format, your TIF images
(microscopy I assume?) will not have such metadata.

So what CMTK is telling you is that it cannot figure out from your
images how their array indexes relate to (human) anatomical standard
orientations. I suspect that's something that simply doesn't apply to
your data.

As a result, CMTK will simply read your images as they are and keep them in memory as they are stored on disk.

If you register two images that have identical gross orientation, then
that's not a problem, because whether or nor they both get reoriented
doesn't change their relative orientation with respect to each other.

When you have different orientations, however, that is, when your images
require something like 90, 180, 270 degree rotations for gross
alignment, then you're in trouble, because CMTK's registration will
typically fail to capture these misorientations.

One solution in this case would be to use the make_initial_affine
command line tool to generate an initial affine transformation based on
Principal Axes ("--principal-axes") and use the result as the input to
"registration". The more recent, extended "registrationx" tool can also
directly do that if you give it the "--pax" command line option.

Trouble is, that principal axes may well fail completely, for example
when your images are truncated on most sides of the field of view
(unlike, say human head images, which have only a small truncation
surface that cuts through the neck).

Another solution would be to manually create an initial affine
transformation. There is, unfortunately, no tool that makes this easy.

The best solution would perhaps be to define a consistent coordinate
system for your imaging process and somehow map it to "RAS" coordinates,
i.e., pretend you are imaging something that is related to human
anatomical coordinates.

I'll be happy to discuss these things further (but perhaps on NITRC?)



On 02/02/2012 04:53 PM,  wrote:

I typically acquire 3d tif stacks and save them into the nrrd format
using Fiji. Sometimes they will register just fine, but recently I
have encountered a couple that give me the following error:
WARNING: image does not have valid orientation meta information; cannot reorient
I thought about opening the files with triplanar, but I cannot launch
that. I also looked online to see if I could find a thread that
addressed this issue and failed to find one. Do you know what I could
Torsten Rohlfing, PhD SRI International, Neuroscience Program
Senior Research Scientist 333 Ravenswood Ave, Menlo Park, CA 94025
Phone: ++1 (650) 859-3379 Fax: ++1 (650) 859-2743

"Though this be madness, yet there is a method in't"
Feb 2, 2012  09:02 PM | Ruben Portugues
RE: image does not have orientation meta informat
Thanks for the reply Torsten, that was extremely useful.  My images are always roughly in the same orientation (+- 10 degrees) so I just proceeded with the warping without paying much attention to the warning message.  After running the reformatx command I am now getting a totally black image, which I guess may be because I have not chosen the correct parameters when warping.  I will fiddle a bit more with this and post again if I continue to have problems.
Feb 3, 2012  09:02 AM | Torsten Rohlfing - Google LLC
RE: image does not have orientation meta informat
Is there any reason why you go straight for the warping rather than first do a linear (affine) registration? (Assuming I understand your reply correctly).

Also, I'll be happy to take a look at your warping parameters if you like.