Image references from SR - DICOM

This is a discussion on Image references from SR - DICOM ; Hi All, In David Clunie's book "DICOM Structured Reporting" and also at this link: http://groups-beta.google.com/group/...7e6fa3924e7321 it is mentioned that basic text SR cannot Image references, but according to PS3.3 - 2004 A.35.1.3.1.1 & A.35.1.3.1.1 Basic text SR can contain IMAGE ...

+ Reply to Thread
Results 1 to 5 of 5

Thread: Image references from SR

  1. Image references from SR

    Hi All,

    In David Clunie's book "DICOM Structured Reporting" and also at this
    link:
    http://groups-beta.google.com/group/...7e6fa3924e7321

    it is mentioned that basic text SR cannot Image references, but
    according to PS3.3 - 2004 A.35.1.3.1.1 & A.35.1.3.1.1 Basic text SR
    can contain IMAGE value types and TEXT value types can be inferred from
    IMAGE value types. I also looked back into previous DICOM standard
    versions (2000) but I didn't see any changes on this matter.
    Furthermore TID 2001 which is mentioned in the above link has inferred
    from IMAGE so I see no reason why the IOD can't be a legitimate Basic
    Text SR if it has an inferred from IMAGE relationship.


    Second question related also to image references is how should one
    retrieve from an archive an image referenced by a SR if the IMAGE value
    type uses the "Referenced SOP Sequence" (0008,1199) which contains only
    SOP class & SOP instance? Is it possible that writer's intention was
    that the report viewer should query the archive until it will find the
    requested SOP instance??
    Should I add always (although it an 1C tag) the tag "Current Requested
    Procedure Evidence Sequence" (0040,a375) in order to keep full track of
    the the referenced images? I haven't seen this sequence to be in use
    (except for Mammo CAD) by any vendor, so even if I add to this tag to
    my SR, I still can't retrive images that were referenced by other
    vendors SR. Any idea how should I proceed ?
    Thanks,
    Leon


  2. Re: Image references from SR

    Hi Leon

    leon wrote:

    > In David Clunie's book "DICOM Structured Reporting" and also at this
    > link:
    > http://groups-beta.google.com/group/...7e6fa3924e7321
    >
    > it is mentioned that basic text SR cannot Image references


    I cannot find anywhere in my book where I say any such thing.

    , but
    > according to PS3.3 - 2004 A.35.1.3.1.1 & A.35.1.3.1.1 Basic text SR
    > can contain IMAGE value types and TEXT value types can be inferred from
    > IMAGE value types. I also looked back into previous DICOM standard
    > versions (2000) but I didn't see any changes on this matter.
    > Furthermore TID 2001 which is mentioned in the above link has inferred
    > from IMAGE so I see no reason why the IOD can't be a legitimate Basic
    > Text SR if it has an inferred from IMAGE relationship.


    A TEXT content item INFERRED FROM an IMAGE content item is indeed
    specifically allowed in the Basic Text SR.

    I can't see anything in my book or the post you reference that says
    otherwise. Gilles post mentioned CODE content items being inferred
    from things, and CODE content items are not permitted in Basic Text SR
    except as leaves.

    > Second question related also to image references is how should one
    > retrieve from an archive an image referenced by a SR if the IMAGE value
    > type uses the "Referenced SOP Sequence" (0008,1199) which contains only
    > SOP class & SOP instance? Is it possible that writer's intention was
    > that the report viewer should query the archive until it will find the
    > requested SOP instance?
    > Should I add always (although it an 1C tag) the tag "Current Requested
    > Procedure Evidence Sequence" (0040,a375) in order to keep full track of
    > the the referenced images? I haven't seen this sequence to be in use
    > (except for Mammo CAD) by any vendor, so even if I add to this tag to
    > my SR, I still can't retrive images that were referenced by other
    > vendors SR. Any idea how should I proceed ?


    I think you are saying that for a hierarchical C-MOVE, the Study Instance UID
    and the Series Instance UID are needed in addition to the SOP Instance
    UID. Only the latter is present in the Referenced SOP Sequence.

    The Current Requested Procedure Evidence Sequence is specifically intended
    to factor out this additional information for the purpose that you
    describe (amongst others).

    Unfortunately, the Current Requested Procedure Evidence Sequence is not
    mandatory.

    Rather, it is conditional on a rather hokey condition described as:

    "Required if Composite Objects were acquired in order to satisfy the
    Requested Procedure(s) for which the SR Document is generated and
    Completion Flag (0040,A491) value is COMPLETE. May be present otherwise."

    Further, Current Requested Procedure Evidence Sequence is described as
    follows (C.17.2.3):

    "For a completed SR Document satisfying (i.e., being the final report for)
    the current Requested Procedure(s), this sequence shall list the full set
    of Composite SOP Instances created for the current Requested Procedure(s).
    For other SOP Instances that include the SR Document General Module, this
    sequence shall contain at minimum the set of Composite SOP Instances from
    the current Requested Procedure(s) that are referenced in the content
    tree."

    This definition is a little inconsistent with the condition itself.

    The confusion arises because we were trying to overload two concepts
    in one attribute - what SOP instances the report applied to, and how
    to retrieve them.

    The unfortunate consequence of this is that if one has a Completion Flag
    (0040,A491) value of PARTIAL, then one does not have to send this sequence,
    or if one can somehow claim that there was no request in the first place,
    then one does not have to send this sequence.

    Currently, all SR IODs except Key Object Selection require the SR Document
    General Module.

    The Key Object Selection IOD requires that the same information be present,
    it just describes the Current Requested Procedure Evidence Sequence in
    the Key Object Document Module.

    If you are indeed seeing SR instances from any vendor that do not contain
    the appropriate information in Current Requested Procedure Evidence Sequence,
    when they are Key Object Documents or when they are SR objects and
    Completion Flag (0040,A491) value is COMPLETE, then they are probably
    invalid SR objects. Please let me know if this is ever the case.

    However, if they have a Completion Flag (0040,A491) value of PARTIAL, then
    they are unfortunately probably legal.

    I suspect that these type of PARTIAL objects will be those referred to
    as "evidence" documents, such as come from the acquisition devices
    making measurements. It will be extremely unfortunate if it becomes
    common practice not to send the Current Requested Procedure Evidence
    Sequence for evidence documents.

    But I doubt that a CP at this stage to make the sequence just plain
    mandatory would succeed, even though the description of the attribute
    seems to conflict with the condition.

    Best that I can suggest in such circumstances is to do as I think that
    you are suggesting and perform a C-FIND at the Study and then Series
    level for the current Study to try to find them, perhaps filtered by
    Modality at the Series level to keep the number of matches tractable.

    Of course the referenced instances might not be in the same Study as
    the SR itself

    Assuming relational queries are not available.

    David

  3. Re: Image references from SR

    Hi David,

    Thank you very much for your answer.
    I am sorry for my mistake, I confused between two non-related issues.

    Regarding the "Current Requested Procedure Evidence Sequence" tag, I
    haven't encountered it in any SR (except Mammo-CAD) until now.
    For example, the file report13.dcm in DicomScope database uses
    reference to image but not the "Current Requested Procedure Evidence
    Sequence" tag, which is OK since the tag is not mandatory.
    But, since many vendors, including myself, use DicomScope for
    validation, I doubt there are many vendors that use the "Current
    Requested Procedure Evidence Sequence".

    Thanks,
    Leon



    David Clunie wrote:
    > Hi Leon
    >
    > leon wrote:
    >
    > > In David Clunie's book "DICOM Structured Reporting" and also at this
    > > link:
    > > http://groups-beta.google.com/group/...7e6fa3924e7321
    > >
    > > it is mentioned that basic text SR cannot Image references

    >
    > I cannot find anywhere in my book where I say any such thing.
    >
    > , but
    > > according to PS3.3 - 2004 A.35.1.3.1.1 & A.35.1.3.1.1 Basic text SR
    > > can contain IMAGE value types and TEXT value types can be inferred from
    > > IMAGE value types. I also looked back into previous DICOM standard
    > > versions (2000) but I didn't see any changes on this matter.
    > > Furthermore TID 2001 which is mentioned in the above link has inferred
    > > from IMAGE so I see no reason why the IOD can't be a legitimate Basic
    > > Text SR if it has an inferred from IMAGE relationship.

    >
    > A TEXT content item INFERRED FROM an IMAGE content item is indeed
    > specifically allowed in the Basic Text SR.
    >
    > I can't see anything in my book or the post you reference that says
    > otherwise. Gilles post mentioned CODE content items being inferred
    > from things, and CODE content items are not permitted in Basic Text SR
    > except as leaves.
    >
    > > Second question related also to image references is how should one
    > > retrieve from an archive an image referenced by a SR if the IMAGE value
    > > type uses the "Referenced SOP Sequence" (0008,1199) which contains only
    > > SOP class & SOP instance? Is it possible that writer's intention was
    > > that the report viewer should query the archive until it will find the
    > > requested SOP instance?
    > > Should I add always (although it an 1C tag) the tag "Current Requested
    > > Procedure Evidence Sequence" (0040,a375) in order to keep full track of
    > > the the referenced images? I haven't seen this sequence to be in use
    > > (except for Mammo CAD) by any vendor, so even if I add to this tag to
    > > my SR, I still can't retrive images that were referenced by other
    > > vendors SR. Any idea how should I proceed ?

    >
    > I think you are saying that for a hierarchical C-MOVE, the Study Instance UID
    > and the Series Instance UID are needed in addition to the SOP Instance
    > UID. Only the latter is present in the Referenced SOP Sequence.
    >
    > The Current Requested Procedure Evidence Sequence is specifically intended
    > to factor out this additional information for the purpose that you
    > describe (amongst others).
    >
    > Unfortunately, the Current Requested Procedure Evidence Sequence is not
    > mandatory.
    >
    > Rather, it is conditional on a rather hokey condition described as:
    >
    > "Required if Composite Objects were acquired in order to satisfy the
    > Requested Procedure(s) for which the SR Document is generated and
    > Completion Flag (0040,A491) value is COMPLETE. May be present otherwise."
    >
    > Further, Current Requested Procedure Evidence Sequence is described as
    > follows (C.17.2.3):
    >
    > "For a completed SR Document satisfying (i.e., being the final report for)
    > the current Requested Procedure(s), this sequence shall list the full set
    > of Composite SOP Instances created for the current Requested Procedure(s).
    > For other SOP Instances that include the SR Document General Module, this
    > sequence shall contain at minimum the set of Composite SOP Instances from
    > the current Requested Procedure(s) that are referenced in the content
    > tree."
    >
    > This definition is a little inconsistent with the condition itself.
    >
    > The confusion arises because we were trying to overload two concepts
    > in one attribute - what SOP instances the report applied to, and how
    > to retrieve them.
    >
    > The unfortunate consequence of this is that if one has a Completion Flag
    > (0040,A491) value of PARTIAL, then one does not have to send this sequence,
    > or if one can somehow claim that there was no request in the first place,
    > then one does not have to send this sequence.
    >
    > Currently, all SR IODs except Key Object Selection require the SR Document
    > General Module.
    >
    > The Key Object Selection IOD requires that the same information be present,
    > it just describes the Current Requested Procedure Evidence Sequence in
    > the Key Object Document Module.
    >
    > If you are indeed seeing SR instances from any vendor that do not contain
    > the appropriate information in Current Requested Procedure Evidence Sequence,
    > when they are Key Object Documents or when they are SR objects and
    > Completion Flag (0040,A491) value is COMPLETE, then they are probably
    > invalid SR objects. Please let me know if this is ever the case.
    >
    > However, if they have a Completion Flag (0040,A491) value of PARTIAL, then
    > they are unfortunately probably legal.
    >
    > I suspect that these type of PARTIAL objects will be those referred to
    > as "evidence" documents, such as come from the acquisition devices
    > making measurements. It will be extremely unfortunate if it becomes
    > common practice not to send the Current Requested Procedure Evidence
    > Sequence for evidence documents.
    >
    > But I doubt that a CP at this stage to make the sequence just plain
    > mandatory would succeed, even though the description of the attribute
    > seems to conflict with the condition.
    >
    > Best that I can suggest in such circumstances is to do as I think that
    > you are suggesting and perform a C-FIND at the Study and then Series
    > level for the current Study to try to find them, perhaps filtered by
    > Modality at the Series level to keep the number of matches tractable.
    >
    > Of course the referenced instances might not be in the same Study as
    > the SR itself
    >
    > Assuming relational queries are not available.
    >
    > David



  4. Re: Image references from SR

    leon wrote:

    > For example, the file report13.dcm in DicomScope database uses
    > reference to image but not the "Current Requested Procedure Evidence
    > Sequence" tag, which is OK since the tag is not mandatory.


    The reason why this sequence is not contained in the report is quite
    simple: at the time when we created the report the underlying DCMTK
    did not yet support this sequence. This has changed in the meantime
    and you should be able to create an enhanced version of the report
    using the current DCMTK 3.5.3 as follows:

    dcmsr/tests/mkreport 13

    > But, since many vendors, including myself, use DicomScope for
    > validation, I doubt there are many vendors that use the "Current
    > Requested Procedure Evidence Sequence".


    In order to validate DICOM SR documents I would rather recommend
    DCMcheck or any other DICOM validation tool. The tool dsrdump from
    the DCMTK also has some useful options in order to check SR documents
    for DICOM conformance.

    Regards,
    Joerg Riesmeier
    OFFIS

  5. Re: Image references from SR

    > This has changed in the meantime and you should be able to create
    > an enhanced version of the report using the current DCMTK 3.5.3 as
    > follows:


    I took this occasion to update the package with SR sample documents:
    ftp://dicom.offis.de/pub/dicom/offis...t/srdoc103.zip

    Regards,
    Joerg Riesmeier
    OFFIS


+ Reply to Thread