Extracting measurements from DICOM images - DICOM

This is a discussion on Extracting measurements from DICOM images - DICOM ; Hello, We acquire radiology images and have the need to extract measurements from them. These measurements are overlaid onto the bitmap image of the DICOM file. However, when I try to see if these values are embedded anywhere in the ...

+ Reply to Thread
Results 1 to 3 of 3

Thread: Extracting measurements from DICOM images

  1. Extracting measurements from DICOM images

    Hello,

    We acquire radiology images and have the need to extract measurements
    from them. These measurements are overlaid onto the bitmap image of the
    DICOM file. However, when I try to see if these values are embedded
    anywhere in the DICOM, I cannot find them. I have used DICOM
    Inspector.NET. Is there any other utilities out there that will show
    all DICOM tags and identifiers?

    Thanks in advance

    Mike McWhinney
    elja, Inc.


  2. Re: Extracting measurements from DICOM images


    eljainc wrote:
    > Hello,
    >
    > We acquire radiology images and have the need to extract measurements
    > from them. These measurements are overlaid onto the bitmap image of the
    > DICOM file. However, when I try to see if these values are embedded
    > anywhere in the DICOM, I cannot find them. I have used DICOM
    > Inspector.NET. Is there any other utilities out there that will show
    > all DICOM tags and identifiers?
    >
    > Thanks in advance
    >
    > Mike McWhinney
    > elja, Inc.


    There are multiple different DICOM tag editors/dumpers. That will allow
    you to see this information if it is there and readily accessible. One
    of them them is in the Oldenburg OFFIS DICOM Tookit DCMTK: dcmdump. See
    http://dicom.offis.de/dcmtk.php.en

    However, what you see presented in a DICOM viewing application is not
    neccessarily fully encoded entirely (or only in) the standard DICOM
    tags included in the image.

    Some/many PACS vendors encode measurements in private data elements
    which most standards based tag dumpers will simply dump as hex values.
    Others don't even put them in the image at all but instead just keep
    them in their application database. Where they have been encoded as
    private data elements, the reason the tag dumpers won't show the
    information is the data type and or structure of the information in
    private tags is by its nature, private. Private tags are indicated by
    tags with odd-valued group numbers (the first four digits of the 8-hex
    digit dicom attribute tag).

    Measurement information can also be encoded into standard DICOM
    elements in several ways. In some cases the modality will actually
    "burn" the infromation into the pixel values. In other cases it can be
    coded into the upper, unused bits of the pixel data bytes. Burned in
    data is usually pretty obivous since it obscures the underlying pixel
    data and cannot be toggled off. Overlay data is also usually pretty
    obvious because , although it can be toggled off, it is clearly raster
    information which "blows up" when the image is zoomed.

    Another way in which measurement data can be encoded that is both a
    vector graphic form and readable is as graphic annotation sequence.
    These sequences can either be embedded in the image itself (discouraged
    but is done, especially when measurement is performed on the modality)
    or the annotation maybe embedded in a separate Presentation State
    object, which references the annotated image --i.e. it is a stand alone
    DICOM object containing, among other things, a graphic annotation
    sequence and pointer to image which has been annotated and the location
    of the annotation in the image. The association between the
    presentation state object and image is encoded in the DICOM tag data
    and is read by the viewing application which has received both objects.
    The viewing application, through its internal logic, renders the
    annotation on the image pixels. So, while you could find the annotation
    data, you would have to retrace that same process the viewing
    application used: find the image file, identify the presentation state
    object which references it, then parse/dump the presentation state
    object to find the graphic annotations (assuming this technique has
    been used).

    Yet another way measurements can be encoded are in DICOM structured
    reports, again through a reference to the image, and an encoded
    description of the measurement.

    I mention this latter approach because of the limitations of the
    presentation state graphic annotation encoding. It is intended to
    encode presentation oriented data --ie. data indicating "Draw a line
    from Pixel (x1,y1) to pixel (x2,y2)" or "draw a circle with 23 pixel
    radius centered on (x3,y3)" and "draw text '4 mm'" at pixel (x4,y4),
    etc. It does not encode information which says "4 mm length
    measurement" or "4 mm diameter mass" at some specified location. The
    graphic annotation approach does not encode any semantics to the
    graphic annotations. Literally, the interpretation is left to the
    viewer (person) who sees the graphics as they are rendered in the
    image. If you're looking to search for, and digitally extract the
    semantics of a measurement in the image, the graphic annotation
    sequence isn't going to give it too you. DICOM structured reports do
    provide this capability but they tend not to be widely supported where
    their measurements are rendered and displayed with the image data
    (outside of some relatively narrow domain implementations such as
    viewing applications for echocardiography , ultrasound, and
    mammography).

    This subject has come up previously. David Clunie had mentioned there
    was work in progress to add the ability to capture additional semantics
    in image annotations. My interpretation of his comment at the time was
    for a sort of light-weight , mini-Structured Report construct which
    could be embedded into a presentation state object or directly into an
    image for that matter. This work may have proceeded and incorporated
    into the standard (a lot of things come into the standard under the
    radar through DICOM correction proposals). I invite David to comment
    on what might have happened with that work, its status in the standard,
    and support in existing applications


  3. Re: Extracting measurements from DICOM images


    eric.goodall@gmail.com wrote:
    > I invite David to comment
    > on what might have happened with that work, its status in the standard,
    > and support in existing applications

    Receiving the agenda for the October Working Group 6 meeting today, I
    can answer my own question. The new features for encoding measurements
    and other "complex annotations" are part of Supplement 120: Extended
    Presentation States. It is being worked in committee and is targeted
    for release for Public Comment at the conclusion of the October
    meeting.
    Here's the current draft:
    ftp://medical.nema.org/medical/dicom.../sup120_08.pdf


+ Reply to Thread