Bsps and multiple modality transforms. - DICOM

This is a discussion on Bsps and multiple modality transforms. - DICOM ; So we have this blending softcopy presentation state that is going to describe how to present a blending of a CT series with a PET series. The presentation state blending module allows for one modality transform for the underlying CT ...

+ Reply to Thread
Results 1 to 5 of 5

Thread: Bsps and multiple modality transforms.

  1. Bsps and multiple modality transforms.

    So we have this blending softcopy presentation state that is going to
    describe how to present a blending of a CT series with a PET series.

    The presentation state blending module allows for one modality
    transform for the underlying CT images and one modality transform for
    the superimposed PET images.

    The problem is that some PET modalities produces a different value for
    rescale-slope for each image in the series, and the bsps only allows a
    single common value for all the images referred by the bsps.

    This leaves us with a nasty choice:
    - We may ignore the modality transform of the bsps in violation of
    the
    standard.
    - We may use the combined modality transform of the image and the
    bsps. Different violation.
    - We may generate multiple bspses. - one per pet-image. Leaving to
    the image display the impossible task of guessing which bspses are
    related.

    Not to mention that we will also have to guess what other vendors did
    to deal with the same problem.

    Hopefully there is something in the standard that I didnt notice.

  2. Re: Bsps and multiple modality transforms.

    On Feb 12, 7:31*pm, j...@medical-insight.com wrote:
    > So we have this blending softcopy presentation state that is going to
    > describe how to present a blending of a CT series with a PET series.
    >
    > The presentation state blending module allows for one modality
    > transform for the underlying CT images and one modality transform for
    > the superimposed PET images.
    >
    > The problem is that some PET modalities produces a different value for
    > rescale-slope for each image in the series, and the bsps only allows a
    > single common value for all the images referred by the bsps.
    >
    > This leaves us with a nasty choice:
    > * - We may ignore the modality transform of the bsps in violation of
    > the
    > * * standard.
    > * - We may use the combined modality transform of the image and the
    > * * bsps. *Different violation.
    > * - We may generate multiple bspses. - one per pet-image. Leaving to
    > * * the image display the impossible task of guessing which bspses are
    > * * related.
    >
    > Not to mention that we will also have to guess what other vendors did
    > to deal with the same problem.
    >
    > Hopefully there is something in the standard that I didnt notice.


    Hi,

    Whatever I can put together is here....If I am wrong or making mistake
    at any point please correct me, this is appeal to all!


    Please see the text from PDF 3 of 2008 Standard.
    "
    C.11 LOOK UP TABLES
    C.11.1 Modality LUT Module
    Table C.11-1 specifies the Attributes that describe the Modality LUT.
    Either a Modality LUT Sequence containing a single Item or Rescale
    Slope and Intercept values
    shall be present but not both.
    Note: This requirement for only a single transformation makes it
    possible to unambiguously define the
    input of succeeding stages of the grayscale pipeline such as the VOI
    LUT.
    "
    I guess the NOTE mentioned here has some key in it.

    One query from my side, you said about different values for Rescale
    slope and intercept for each PET image in a series.
    But this information will be present in the Image only, correct.
    And as per PDF 4 Annex N "Section N.2.1.1 Modality LUT" the
    transformation information from Image should not be used when we are
    applying any Presentation State, even though the same transformation
    information is not present in Presentation State which is nothing but
    IDENTITY Transformation Case.



    -Regards,
    Prashant




  3. Re: Bsps and multiple modality transforms.

    On Feb 13, 2:26 am, PacificPrashant wrote:
    > On Feb 12, 7:31 pm, j...@medical-insight.com wrote:
    >
    >
    >
    > > So we have this blending softcopy presentation state that is going to
    > > describe how to present a blending of a CT series with a PET series.

    >
    > > The presentation state blending module allows for one modality
    > > transform for the underlying CT images and one modality transform for
    > > the superimposed PET images.

    >
    > > The problem is that some PET modalities produces a different value for
    > > rescale-slope for each image in the series, and the bsps only allows a
    > > single common value for all the images referred by the bsps.

    >
    > > This leaves us with a nasty choice:
    > > - We may ignore the modality transform of the bsps in violation of
    > > the
    > > standard.
    > > - We may use the combined modality transform of the image and the
    > > bsps. Different violation.
    > > - We may generate multiple bspses. - one per pet-image. Leaving to
    > > the image display the impossible task of guessing which bspses are
    > > related.

    >
    > > Not to mention that we will also have to guess what other vendors did
    > > to deal with the same problem.

    >
    > > Hopefully there is something in the standard that I didnt notice.

    >
    > Hi,
    >
    > Whatever I can put together is here....If I am wrong or making mistake
    > at any point please correct me, this is appeal to all!
    >
    > Please see the text from PDF 3 of 2008 Standard.
    > "
    > C.11 LOOK UP TABLES
    > C.11.1 Modality LUT Module
    > Table C.11-1 specifies the Attributes that describe the Modality LUT.
    > Either a Modality LUT Sequence containing a single Item or Rescale
    > Slope and Intercept values
    > shall be present but not both.
    > Note: This requirement for only a single transformation makes it
    > possible to unambiguously define the
    > input of succeeding stages of the grayscale pipeline such as the VOI
    > LUT.
    > "
    > I guess the NOTE mentioned here has some key in it.
    >
    > One query from my side, you said about different values for Rescale
    > slope and intercept for each PET image in a series.
    > But this information will be present in the Image only, correct.
    > And as per PDF 4 Annex N "Section N.2.1.1 Modality LUT" the
    > transformation information from Image should not be used when we are
    > applying any Presentation State, even though the same transformation
    > information is not present in Presentation State which is nothing but
    > IDENTITY Transformation Case.


    My problem is that I have a number of pet images in a series, each
    image with a different slope.

    I need to create a bsps describing how to render this series, and I
    want to use the slope that is already in the image because that is the
    transform that brings the pixel values into a common value space.

    But the bsps encoding only allows me to encode a single common value
    for all the images. Yes it is unambigious, but it is also wrong.

    If this had been a gsps I would simply create a separate gsps for each
    image and all is fine. - But that is not an option with a bsps because
    the rendering of a bsps may require resampling from multiple images in
    a series, which means that I *must* have the pixel values in the same
    space in a single bsps.

    What to do?

    What do other vendors do ?

    Currently I think my best bet is to detect the situation, and then use
    the modality transform from the image rather than from the bsps. While
    this is a violation of the standard it will at least produce a proper
    rendering as long as the images where recorded correctly in the first
    place.

    Of course the whole problem goes away when the the images are
    multiframe (enhanced pet-image storage class) because that will only
    allow a single slope on the frames.

    >
    > -Regards,
    > Prashant



  4. Re: Bsps and multiple modality transforms.

    On Feb 13, 6:02*pm, j...@medical-insight.com wrote:
    > On Feb 13, 2:26 am, PacificPrashant wrote:
    >
    >
    >
    >
    >
    > > On Feb 12, 7:31 pm, j...@medical-insight.com wrote:

    >
    > > > So we have this blending softcopy presentation state that is going to
    > > > describe how to present a blending of a CT series with a PET series.

    >
    > > > The presentation state blending module allows for one modality
    > > > transform for the underlying CT images and one modality transform for
    > > > the superimposed PET images.

    >
    > > > The problem is that some PET modalities produces a different value for
    > > > rescale-slope for each image in the series, and the bsps only allows a
    > > > single common value for all the images referred by the bsps.

    >
    > > > This leaves us with a nasty choice:
    > > > * - We may ignore the modality transform of the bsps in violation of
    > > > the
    > > > * * standard.
    > > > * - We may use the combined modality transform of the image and the
    > > > * * bsps. *Different violation.
    > > > * - We may generate multiple bspses. - one per pet-image. Leaving to
    > > > * * the image display the impossible task of guessing which bspsesare
    > > > * * related.

    >
    > > > Not to mention that we will also have to guess what other vendors did
    > > > to deal with the same problem.

    >
    > > > Hopefully there is something in the standard that I didnt notice.

    >
    > > Hi,

    >
    > > Whatever I can put together is here....If I am wrong or making mistake
    > > at any point please correct me, this is appeal to all!

    >
    > > Please see the text from PDF 3 of 2008 Standard.
    > > "
    > > C.11 LOOK UP TABLES
    > > C.11.1 Modality LUT Module
    > > Table C.11-1 specifies the Attributes that describe the Modality LUT.
    > > Either a Modality LUT Sequence containing a single Item or Rescale
    > > Slope and Intercept values
    > > shall be present but not both.
    > > Note: This requirement for only a single transformation makes it
    > > possible to unambiguously define the
    > > input of succeeding stages of the grayscale pipeline such as the VOI
    > > LUT.
    > > "
    > > I guess the NOTE mentioned here has some key in it.

    >
    > > One query from my side, you said about different values for Rescale
    > > slope and intercept for each PET image in a series.
    > > But this information will be present in the Image only, correct.
    > > And as per PDF 4 Annex N "Section N.2.1.1 Modality LUT" the
    > > transformation information from Image should not be used when we are
    > > applying any Presentation State, even though the same transformation
    > > information is not present in Presentation State which is nothing but
    > > IDENTITY Transformation Case.

    >
    > My problem is that I have a number of pet images in a series, each
    > image with a different slope.
    >
    > I need to create a bsps describing how to render this series, and I
    > want to use the slope that is already in the image because that is the
    > transform that brings the pixel values into a common value space.
    >
    > But the bsps encoding only allows me to encode a single common value
    > for all the images. Yes it is unambigious, but it is also wrong.
    >
    > If this had been a gsps I would simply create a separate gsps for each
    > image and all is fine. - But that is not an option with a bsps because
    > the rendering of a bsps may require resampling from multiple images in
    > a series, which means that I *must* have the pixel values in the same
    > space in a single bsps.
    >
    > What to do?
    >
    > What do other vendors do ?
    >
    > Currently I think my best bet is to detect the situation, and then use
    > the modality transform from the image rather than from the bsps. While
    > this is a violation of the standard it will at least produce a proper
    > rendering as long as the images where recorded correctly in the first
    > place.
    >
    > Of course the whole problem goes away when the the images are
    > multiframe (enhanced pet-image storage class) because that will only
    > allow a single slope on the frames.
    >
    >
    >
    >
    >
    > > -Regards,
    > > Prashant- Hide quoted text -

    >
    > - Show quoted text -- Hide quoted text -
    >
    > - Show quoted text -


    Hi,

    Not getting any clue...yet!
    Let's wait for some Experts Advice...
    Helloooo......Anybody reading this?

    Regards,
    Prashant


  5. Re: Bsps and multiple modality transforms.

    Hi

    BSPS, like all existing presentation states, allows for only a
    single pair of Rescale Slope/Intercept values that must apply
    to all the reference images (or more specifically, each set of
    underlying or superimposed images).

    Further, BSPS, like all presentation states, requires that any
    transformations in the referenced images be overridden by the
    transformations in the presentation state itself. I.e., the
    presentation state defines the entire pipeline, and a missing
    set of values means an identity transformation, not to use what
    is define for that step in the images.

    As far as I know there is only one vendor that insists on sending
    a different Rescale Slope per PET slice, and that is GE. They say
    that given a fixed bit depth they want to make maximum use of the
    dynamic range for each slice (e.g., there are a lot more counts
    in the bladder than the middle of the abdomen, etc.).

    Anyway, this means that given a set of images in a PET Series
    that have such a pattern of varying Rescale Slopes, one must
    do something creative in a BSPS that references the entire set.

    One cannot have a separate BSPS per image, since the entire set
    needs to be referenced from one BSPS to convey that they are a
    set to resampled together.

    What GE does when creating a BSPS for their images, from the
    samples I have seen, is to factor out a common Rescale Slope,
    define the Rescale Type to be US and to send fractional Window
    Center/Width values adjusted accordingly such that when the rescaling
    and windowing are applied, the grayscale intensity is consistent
    from slice to slice. This is possible since the rescale
    (modality LUT) step and the windowing (VOI LUT) step are just two
    successive linear operations.

    Note that to report SUV or activity/concentration (MBq/ml) from an
    ROI, one needs to use the information from the original image's
    Rescale Slope and Units attributes, not the common Rescale Slope
    sent in such a BSPS.

    David

    jbb@medical-insight.com wrote:
    > So we have this blending softcopy presentation state that is going to
    > describe how to present a blending of a CT series with a PET series.
    >
    > The presentation state blending module allows for one modality
    > transform for the underlying CT images and one modality transform for
    > the superimposed PET images.
    >
    > The problem is that some PET modalities produces a different value for
    > rescale-slope for each image in the series, and the bsps only allows a
    > single common value for all the images referred by the bsps.
    >
    > This leaves us with a nasty choice:
    > - We may ignore the modality transform of the bsps in violation of
    > the
    > standard.
    > - We may use the combined modality transform of the image and the
    > bsps. Different violation.
    > - We may generate multiple bspses. - one per pet-image. Leaving to
    > the image display the impossible task of guessing which bspses are
    > related.
    >
    > Not to mention that we will also have to guess what other vendors did
    > to deal with the same problem.
    >
    > Hopefully there is something in the standard that I didnt notice.


+ Reply to Thread