Unique identifier for an image - DICOM

This is a discussion on Unique identifier for an image - DICOM ; What is the best way of uniquely identifying a DICOM image so that I can search my database and pull out a specific image? That is to say, what header info is guaranteed to be unique for each image so ...

+ Reply to Thread
Results 1 to 6 of 6

Thread: Unique identifier for an image

  1. Unique identifier for an image

    What is the best way of uniquely identifying a DICOM image so that I
    can search my database and pull out a specific image? That is to say,
    what header info is guaranteed to be unique for each image so I can use
    it as the primary key in my image table?

    What I need to be able to do is differentiate between one image and
    another even if the differences are just that the second image has some
    slight change to the header information or the image data (e.g.
    annotation) in respect of the first image. How do I generate a new
    unique identifer for this second version of the image? Are there any
    specific recommendations for handling different versions of the same
    image? What other header fields might I need to change for this second
    version?

    I assume that any UPDATE of an image record in the database will result
    in a new record being created containing my changes so that the
    original record remains unchanged. Can I establish a relationship
    between these two images using fields in the DICOM header? How can this
    be done?


  2. Re: Unique identifier for an image


    OSPACS wrote:
    > What is the best way of uniquely identifying a DICOM image so that I
    > can search my database and pull out a specific image? That is to say,
    > what header info is guaranteed to be unique for each image so I can use
    > it as the primary key in my image table?
    >
    > What I need to be able to do is differentiate between one image and
    > another even if the differences are just that the second image has some
    > slight change to the header information or the image data (e.g.
    > annotation) in respect of the first image. How do I generate a new
    > unique identifer for this second version of the image? Are there any
    > specific recommendations for handling different versions of the same
    > image? What other header fields might I need to change for this second
    > version?
    >
    > I assume that any UPDATE of an image record in the database will result
    > in a new record being created containing my changes so that the
    > original record remains unchanged. Can I establish a relationship
    > between these two images using fields in the DICOM header? How can this
    > be done?


    Search this forum for SOP Instance UID or just for "UID"

    Razvan


  3. Re: Unique identifier for an image


    > Search this forum for SOP Instance UID or just for "UID"
    >

    Done that, but it doesn't give me the information I need. It seems
    everyone is hung-up on various aspects of creating UIDs rather than
    giving a clear explanation of what uniqueness actually means in terms
    of an image. Thanks for your help anyway.


  4. Re: Unique identifier for an image

    OSPACS wrote:
    > > Search this forum for SOP Instance UID or just for "UID"
    > >

    > Done that, but it doesn't give me the information I need. It seems
    > everyone is hung-up on various aspects of creating UIDs rather than
    > giving a clear explanation of what uniqueness actually means in terms
    > of an image. Thanks for your help anyway.


    I thought that finding a way of generating UIDs was part of your
    question.

    You probably have already extracted the information that the SOP
    Instance UID is the unique identifier of an image.
    It is a mandatory dicom element and it *should* uniquely identify each
    image in the universe and beyond.

    But,as you mentioned, everyone is hung up on various aspects of
    creating UIDs and the reason there's so much fuss about it is that the
    DICOM standard enforces the uniqueness of these UIDs and therefore
    there's a great, great appeal for programmers and system designers to
    use it as keys in their data models.

    However, the rift between what the standard says and what the people do
    is apparent and you've already found some of these issues being debated
    in various posts.

    As for versions, there is no "native" versioning system in dicom as of
    this moment.
    Each modification to an image is stored in a new image, with a new SOP
    Instance UID.

    The handling of this problem is usually done at system level in this
    case, where metadata information (some, or all of it) is usually stored
    in relational databases on which the modifications are operated -on
    patient demographics, for instance.
    When an object is asked for through one of the dicom interfaces
    (C-MOVE/FIND/GET), it is "reconstructed" from the database data and the
    stored data.

    There are many ways on handling this problem. I would suggest for
    starters to check on some of the existing open source PACS software
    like Conquest or K-Pacs or OFFIS or others. This is an issue with
    lotsa' strings attached.

    HTH.
    Razvan


  5. Re: Unique identifier for an image


    Many thanks for this excellent post. It contains jus the sort of
    practical advice that is missing from the DICOM standard.

    Regards

    Will


  6. Re: Unique identifier for an image

    Hi,
    I would like to suggest use of API functions. Generally all dicom API
    come with a function which generates a UID with some vendor prefix (to
    be obtained from DICOM consortium).

    If you are looking for implementation of generation of UID then here is
    some hint.
    UID is 64 char long. around 12-16 chars are cocnsumed by prefix (unique
    to each vendor/implementation). Rest of the character can be generated
    with some predecided scheme. random number can be used but I dont
    recomend because it does not generate numbers in serial order. I prefer
    haven long format date or unix time. Also, you can include some sort of
    site id to keep UID unique across sites. Something of following sort is
    advisable.
    ..

    Hope it helps you
    Bye


    OSPACS wrote:
    > Many thanks for this excellent post. It contains jus the sort of
    > practical advice that is missing from the DICOM standard.
    >
    > Regards
    >
    > Will



+ Reply to Thread