• Problems with localizers

    From Frank@21:1/5 to All on Mon Mar 22 08:59:41 2021
    Hi everyone
    I have a problem with some localizers. In some series of a CT, the first image is a localizer. It is not possibile to recognize it as a localizer from code since the image type is the same as the other images, there is no description and no comments.
    I want to recognize it some way to perform some operations, maybe with the different positions. What do you think about it? Also, having only the image position and the image orientation, how can I certainily say that an image is a localizer and the
    others are the "normal images" of a series?

    Thank all.

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From David Gobbi@21:1/5 to All on Mon Mar 22 12:15:08 2021
    You won't be able to be "certain", and AFAIK there isn't anything in the DICOM standard about it, but you can still make fairly good predictions with ImagePositionPatient and ImageOrientationPatient.

    You've probably already guessed that "ImageOrientationPatient" is usually the same for all images in the series, with the localizer as an exception. And "ImagePositionPatient" will usually progress along a straight line from the first to the last image,
    after localizers have been removed from the series. You can check the progression of ImagePositionPatient in by examining the images in groups of three, e.g. for images 1,2,3 and then for images 2,3,4 etc. If 1,2,3 does not show progression but 2,3,4
    and 3,4,5 do show progression, then 1 is out and 2,3,4,5,... are in.

    There are several caveats:

    1) The ImageOrientationPatient is sometimes provided by a mechanical directional encoder inside the scanner, and like any measurement device, it will be susceptible to noise. So it is possible for some scanners (though very rare) that the
    ImageOrientationPatient will fluctuate very slightly during the scan. So when you compare the ImageOrientationPatient between images, you must allow a small tolerance.

    2) Sometimes a single series can contain two or three sweeps in different directions. I have often seen MR scan protocols begin with a scout series that contains axial, sagittal, and coronal sweeps all within the same series.

    3) The ImageOrientationPatient can hypothetically be different for every image in the series. In rare cases such as these, you have little hope of figuring out which image is the localizer unless it is clearly labelled as such.

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Frank@21:1/5 to All on Tue Mar 23 00:52:45 2021
    Il giorno lunedì 22 marzo 2021 alle 20:15:10 UTC+1 david...@gmail.com ha scritto:
    You won't be able to be "certain", and AFAIK there isn't anything in the DICOM standard about it, but you can still make fairly good predictions with ImagePositionPatient and ImageOrientationPatient.

    You've probably already guessed that "ImageOrientationPatient" is usually the same for all images in the series, with the localizer as an exception. And "ImagePositionPatient" will usually progress along a straight line from the first to the last image,
    after localizers have been removed from the series. You can check the progression of ImagePositionPatient in by examining the images in groups of three, e.g. for images 1,2,3 and then for images 2,3,4 etc. If 1,2,3 does not show progression but 2,3,4
    and 3,4,5 do show progression, then 1 is out and 2,3,4,5,... are in.

    There are several caveats:

    1) The ImageOrientationPatient is sometimes provided by a mechanical directional encoder inside the scanner, and like any measurement device, it will be susceptible to noise. So it is possible for some scanners (though very rare) that the
    ImageOrientationPatient will fluctuate very slightly during the scan. So when you compare the ImageOrientationPatient between images, you must allow a small tolerance.

    2) Sometimes a single series can contain two or three sweeps in different directions. I have often seen MR scan protocols begin with a scout series that contains axial, sagittal, and coronal sweeps all within the same series.

    3) The ImageOrientationPatient can hypothetically be different for every image in the series. In rare cases such as these, you have little hope of figuring out which image is the localizer unless it is clearly labelled as such.
    Hi and thank you very much David. As I thought this is a hard problem. I will certainly give your solution a try.

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)