Stacking order of multi-frame image slices - DICOM

This is a discussion on Stacking order of multi-frame image slices - DICOM ; Hi there! I am having some trouble determining the order of the slices of volumetric multi-frame images. Let's assume the image is transaxial in standard orientation, that is Image Orientation (Patient) is (1,0,0)/(0,1,0). In case of enhanced CT/MR (modalities I ...

+ Reply to Thread
Results 1 to 3 of 3

Thread: Stacking order of multi-frame image slices

  1. Stacking order of multi-frame image slices

    Hi there!

    I am having some trouble determining the order of the slices of volumetric
    multi-frame images.

    Let's assume the image is transaxial in standard orientation, that is Image
    Orientation (Patient) is (1,0,0)/(0,1,0).

    In case of enhanced CT/MR (modalities I do not know much of yet) I think I
    have the Image Position (Patient) for each slice, and I would use the z
    component for ordering just as in the case of single-sliced DICOM files. No
    problem there.

    For other files, I would look if there is a slice vector, and re-order the
    slices accordingly.

    Then, the stacking order should go from head to feet (in my case of a
    transaxial image), or contrary to the direction of the cross product of the
    Image Position (Patient) vectors. Right?

    That is, unless the the Spacing Between Slices is negative, in which case
    the order is reversed. At least this is what the DICOM standard says about
    NM images. What about CT and MR images? I think multi-frame images of these
    modalities are uncommon or even not allowed, but I happen to have some of
    those. Some, having similar DICOM tags, have contrary stacking order, but I
    am not sure which one is correct.
    Same goes for my few volumetric multi-frame NM images. Using the algorithm I
    desribed above, half of them show up correct, the other half is mirrored.

    So, if someone could elaborate on this, I would be very thankful. I also
    would like pointers to sample images. I found only few multi-frame DICOM
    sample images, most of them being not volumetric but time series images,
    and the rest is either sagittal or shows blurry kidneys, so I cannot see
    which ordering would be correct.

    Alex

  2. Re: Stacking order of multi-frame image slices

    Alex,

    On Aug 5, 1:40*pm, Alex Schuster wrote:
    > Hi there!
    >
    > I am having some trouble determining the order of the slices of volumetric
    > multi-frame images.
    >
    > Let's assume the image is transaxial in standard orientation, that is Image
    > Orientation (Patient) is (1,0,0)/(0,1,0).
    >
    > In case of enhanced CT/MR (modalities I do not know much of yet) I think I
    > have the Image Position (Patient) for each slice, and I would use the z
    > component for ordering just as in the case of single-sliced DICOM files. No
    > problem there.
    >
    > For other files, I would look if there is a slice vector, and re-order the
    > slices accordingly.
    >
    > Then, the stacking order should go from head to feet (in my case of a
    > transaxial image), or contrary to the direction of the cross product of the
    > Image Position (Patient) vectors. Right?
    >
    > That is, unless the the Spacing Between Slices is negative, in which case
    > the order is reversed. At least this is what the DICOM standard says about
    > NM images. What about CT and MR images? I think multi-frame images of these
    > modalities are uncommon or even not allowed, but I happen to have some of
    > those. Some, having similar DICOM tags, have contrary stacking order, butI
    > am not sure which one is correct.
    > Same goes for my few volumetric multi-frame NM images. Using the algorithm I
    > desribed above, half of them show up correct, the other half is mirrored.


    IMHO I do not think there is an answer for your question, at least not
    in this dicom newsgroup. After all we are talking about the DICOM
    standard. So here is my opinion.

    If some 3rd party vendor decide to implement extended SOP class, they
    have to produce some kind of DICOM Conformance Statement. You should
    contact them directly.

    A couple of days ago I found one of those files:
    http://groups.google.com/group/comp....3cbfedef33f725
    I thought this would be extremely rare...

    I think that what you assumed is somewhat correct, I mean worse
    scenario is that you have the slices inverted.
    In a clinical trial environment (during site qualification I guess),
    you need to check again with your site, to adjust their machine to
    produce proper DICOM files.

    > So, if someone could elaborate on this, I would be very thankful. I also
    > would like pointers to sample images. I found only few multi-frame DICOM
    > sample images,


    Well, thankfully ! Any vendor producing those things should be
    contacted very quickly so that they can produce a software update that
    fix this issue for the rest of us.

    > most of them being not volumetric but time series images,
    > and the rest is either sagittal or shows blurry kidneys, so I cannot see
    > which ordering would be correct.


    2cts
    -Mathieu

  3. Re: Stacking order of multi-frame image slices

    Mathieu Malaterre writes:

    > On Aug 5, 1:40*pm, Alex Schuster wrote:


    >> I am having some trouble determining the order of the slices of
    >> volumetric multi-frame images.

    [...]

    > IMHO I do not think there is an answer for your question, at least not
    > in this dicom newsgroup. After all we are talking about the DICOM
    > standard. So here is my opinion.
    >
    > If some 3rd party vendor decide to implement extended SOP class, they
    > have to produce some kind of DICOM Conformance Statement. You should
    > contact them directly.


    Okay, thanks for confirming this. So in case of multi-frame, non-NM data,
    the direction is just undefined.

    > A couple of days ago I found one of those files:
    >

    http://groups.google.com/group/comp....3cbfedef33f725
    > I thought this would be extremely rare...


    Ouch.

    > I think that what you assumed is somewhat correct, I mean worse
    > scenario is that you have the slices inverted.


    Yes, and this must not happen. Especially with sagittal slices, there is no
    visual confirmation of the correctness of the order of slices.

    > In a clinical trial environment (during site qualification I guess),
    > you need to check again with your site, to adjust their machine to
    > produce proper DICOM files.


    Seems like this will be the way to go.

    For my strange MR and CT data, I am not surprised that the order is
    arbitrary. But for NM images, the DICOM standard is clear I think. Still I
    have images of different order - unless I just ignore the sign of the
    Spacing Between Slices.
    When I view these images in Osirix, they are in the same order - but all
    upside down.


    >> So, if someone could elaborate on this, I would be very thankful. I also
    >> would like pointers to sample images. I found only few multi-frame DICOM
    >> sample images,

    >
    > Well, thankfully ! Any vendor producing those things should be
    > contacted very quickly so that they can produce a software update that
    > fix this issue for the rest of us.


    What's wrong about multi-frame? AFAIK, NM images _must_ be multi-frame, even
    if they have one slice only. And I am still looking for a correct sample
    file to confirm my viewer gets the slice order right.

    Alex

+ Reply to Thread