Pixel spacing test images for CR/DX/MG - DICOM

This is a discussion on Pixel spacing test images for CR/DX/MG - DICOM ; Hi I have created a set of test images to check for problems with pixel spacing related attributes when making distance and size measurements in PACS systems. Display each image, measure the bar and compare the measurements with the instructions ...

+ Reply to Thread
Results 1 to 5 of 5

Thread: Pixel spacing test images for CR/DX/MG

  1. Pixel spacing test images for CR/DX/MG

    Hi

    I have created a set of test images to check for problems
    with pixel spacing related attributes when making distance
    and size measurements in PACS systems.

    Display each image, measure the bar and compare the measurements
    with the instructions burned into each image.

    See:

    http://www.dclunie.com/images/pixels...testimages.zip

    You may also be able to use these to test any true-size printing
    feature and see if the distance you measure on the print matches
    the instructions.

    David


  2. Re: Pixel spacing test images for CR/DX/MG

    Hi David,

    thanks for the images, you say, that measuring on a CR by using
    0028,0030 would be wrong, but is the error not already the inclusion of
    this element in a CR image, as the image plane module is not part of
    the CR IMAGE IOD?

    By the way, your CR image has a modality value of DX

    After getting curious what the different modality vendors are entering
    I found a totally different picture:

    Most are setting the Pixel Spacing for CRs, some are setting both like
    AGFA and some are only providing the Imager Pixel Spacing.

    So rigorously stating that using the Pixel Spacing on CRs for
    measurement is wrong seem to me inappropriate.

    What would be your approach if both elements are present in an image?
    My approach would be to give the Pixel Spacing a higher priority as the
    information for the user (physician) is of higher interest than the
    Imager Pixel Spacing.


  3. Re: Pixel spacing test images for CR/DX/MG

    coladrops wrote:

    > thanks for the images, you say, that measuring on a CR by using
    > 0028,0030 would be wrong, but is the error not already the inclusion of
    > this element in a CR image, as the image plane module is not part of
    > the CR IMAGE IOD?


    The Image Plane Module (and the corresponding Pixel Spacing attribute)
    are deliberately not part of the CR Image IOD, since that attribute is
    defined to be the spacing in the patient, something that is not known
    for projection radiography modalities, due to the diverging x-ray beam.

    Imager Pixel Spacing, on the other hand, is defined to be equivalent to
    what one would measure if using film instead of CR or a digital sensor,
    and has meaning at the front of the detector or sensor. This is the
    attribute defined by the standard for use in CR, DX, MG, XA and XRF.

    > By the way, your CR image has a modality value of DX


    Thanks, I will change that.

    > After getting curious what the different modality vendors are entering
    > I found a totally different picture:
    >
    > Most are setting the Pixel Spacing for CRs


    And they are all wrong.

    >, some are setting both like AGFA


    Which is tolerable, if both Image Pixel Spacing and Pixel Spacing are
    set to the same value.

    > and some are only providing the Imager Pixel Spacing.


    Which is correct.

    > So rigorously stating that using the Pixel Spacing on CRs for
    > measurement is wrong seem to me inappropriate.


    No, it is completely appropriate ... the standard does not specify the
    meaning of Pixel Spacing for CR (or DX) ... many of the modality vendors
    are doing the wrong thing.

    > What would be your approach if both elements are present in an image?
    > My approach would be to give the Pixel Spacing a higher priority as the
    > information for the user (physician) is of higher interest than the
    > Imager Pixel Spacing.


    And this you have completely around the wrong way, since anyone who makes
    measurements from a projection radiography modality will be assuming that
    they are equivalent to what they would measure on film, unless informed
    otherwise.

    If both attributes are present, then at present, the only attribute that
    has a meaning that is defined by the standard is Imager Pixel Spacing,
    and that should take priority.

    If Pixel Spacing is present, it has no defined meaning, since its
    use in these IODs is not defined by the standard.

    The practical implication of this, is when Pixel Spacing is present in
    addition to or instead of Imager Pixel Spacing, but has a value that
    would be equivalent to what would be sent in Imager Pixel Spacing, then
    there is no problem, since the measurements come out the same (other than
    that one can never be 100% sure that is indeed the meaning as intended
    by the vendor).

    The big problem occurs when both Pixel Spacing and Imager Pixel Spacing
    are present and they have different values !

    This may have resulted in an incident in which it appears that the PACS
    software, whilst doing true size printing, used the wrong attribute, the
    orthopedic surgeon made his measurements to choose a prosthesis for a hip
    replacement, and the result was a patient with one leg significantly shorter
    than the other leg.

    Why would Pixel Spacing have different values than Imager Pixel Spacing ?

    Either a result of a bug (and this has happened, but been corrected by
    the vendor with a recall), or because the modality vendor has creatively
    used Pixel Spacing to somehow "calibrate" for the geometric magnification
    going on during the acquisition. But since the effect of this is
    unpredictable, and as far as I can tell, undocumented, and it is
    potentially unsafe unless the user is aware of it.

    Regardless, given this mess in the installed base, it is now vital that
    users play close attention to whether or not their PACS or workstation
    uses the standard attribute Imager Pixel Spacing, or not, and if when
    both are present, it gives the user the option of one or the other or
    both, for distance measurement and true-size printing, indicating perhaps
    that one is the equivalent of film and the other is "calibrated".

    I have proposed adding a CP to the DICOM standard that defines the
    additional optional use of the Pixel Spacing attribute and that when
    it differs from Imager Pixel Spacing that it is calibrated. This is
    rather closing the barn door after the horse has bolted, but at least
    it would provide a place to document this issue and its significance.

    Regardless, the incident to the patient, which not only has a huge
    impact on that patient, as well as a tremendous medico-legal and product
    liability potential, should serve as a reminder to those who are keen
    to be creative in their interpretation of the standard, or pragmatic
    in their implementation, or allow everything to be totally configurable
    without adequately educating the field engineers and the operators, that
    these are still medical devices and that the potential to do harm to
    patients is significant.

    David

    PS. Just to summarize what I propose might be a safe procedure given
    the installed base of modalities:

    If Imager Pixel Spacing and Pixel Spacing are present:

    - if they have the same values, there is no problem,

    - if they have different values, then Imager Pixel Spacing should represent
    the closest to a film-like measurement, and Pixel Spacing may be some
    sort of calibrated value - let the user decide on a case by case basis
    here, informing them of the implication; the one exception is one vendor's
    CR bug that may be difficult to detect, but if they are a factor of
    two different, and Image Pixel Spacing is less than Pixel Spacing (would
    be the other way around if correcting for geometric magnification),
    and multiplying by rows and columns doesn't match a known
    CR cassette size, you may be able to detect this; though this should have
    been fixed in recalled and new versions, there may be priors in the
    archive with the wrong value, or devices that have not been fixed.

    If Imager Pixel Spacing alone is present use it in the intended manner.

    If Pixel Spacing alone is present, use it, but warn the user every time that
    its meaning is uncertain.

    It may be necessary to require that whenever Pixel Spacing is used instead
    of Imager Pixel Spacing that the screen and/or printed film display a warning
    that measurements may not be true-size in the sense of equivalent to a
    measurement made from film-screen radiography.

    Note also that the display and printing software really should be making its
    decisions based on the image SOP Class, and not ignoring the SOP Class and
    using the same attributes for all images regardless of the IOD defined for
    that SOP Class.

    I also believe it would also be desirable for CR vendors to stop sending
    Pixel Spacing, both to reduce the risk of an error of this type, and to
    force those PACS vendors who do not support Imager Pixel Spacing at all
    to do so if they want to support measurements and true-size printing.

  4. Re: Pixel spacing test images for CR/DX/MG

    Hi David,

    thanks for your detailed answer, I totally agree, that this is a safety
    relevant issue.
    It's very good, that you are putting this on the agenda of the DICOM
    committee.


  5. Re: Pixel spacing test images for CR/DX/MG

    David,

    I have been reviewing DICOM PS 3 - 2007 and it appears the committee has taken many of your recomendations to heart. In specific PS 3.3 section 10.7 now includes the "Basic Pixel Spacing Calibration Macro". That macro has been included in the following Modules.

    CR Image Module C.8.1.2
    SC Image Module C.8.6.2
    SC Multi-Frame Image Module C.8.6.3
    X-Ray Acquisition Module C.8.7.2
    DX Detector Module C.8.11.4

    Based on these modules and Table A.1-1 the Calibration Macro now applies to the following IODs CR, SC, SC-MF-SB, SC-MF-GB, SC-MF-GW, SC-MF-TC, RF, DX, MG, IO.

    The Pixel Spacing value (0028,0030) is included in the "Basic Pixel Spacing Calibration Macro" as well as the "Image Plane Module". My reading of the macro indicates in this context Pixel Spacing now specifically and umambiguosly states it represents the distance in the patient for measurements. I also read it to mean when it is different from Image Pixel Spacing (0018,1164) then it should be interpreted as a correction to account for either geometric magnification or calibration against a fiducial of known size within the image.

    The spec goes further to say Imager Pixel Spacing (0018, 1164) may not be modified to account for any type of magnification or calibration.

    In light of these additions to the specification shouldn't the expected results of the tests that include both Pixel Spacing (0028, 0030) and Imager Pixel Spacing (0018,1164) be changed to use the Pixel Spacing value to calculate the preferred result?

    I am assuming Pixel Spacing, when different from Imager Pixel Spacing, would now be preferred due to the fact it is a "calibrated" value. It should not suffer as badly from errors produced in both film and digital media by geometric magnification. Because DICOM prohibits Imager Pixel Spacing from being calibrated Pixel Spacing has been specifically tagged as the method of providing calibrated values.

    I would appreciate your thoughts.

    Al Meyer

+ Reply to Thread