I am working on DICOM images of chest X ray images. In this, when I load the image in DICOM viewer, the images are having good contrast ratio.
And when I load the same DICOM in python using pydicom or SimpleITK, and view them. They look different than dicom viewer. Is there any intermediate step that I am missing, which is done after loading the DICOM as numpy array?
Gunjan Naik
My best guess is that the DICOM viewer performs some sort of histogram equalization to improve contrast. You may try to perform that operation after loading your data in Python. Please see: https://stackoverflow.com/questions/28518684/histogram-equalization-of-grayscale-images-with-numpy
The problem can be related to the Presentation State of the DICOM images.
A Presentation State is an independent DICOM SOP Instance that contains information on how a particular image should be displayed. The Presentation State may contain label information(types of Label and Positions), windowing values, zoom value, scrolling (panning) values, rotations or any other visual display element that is defined within the DICOM standard.
However Presentation States contain no Pixel Data as they are intended for use with an existing Dicom Image. A presentation state can be applied to an image so that the image is displayed with all the visual specifications defined by that presentation state. The advantage of using presentation states is that you can always revert back to the original image as the underlying pixel data is not modified but rather displayed differently. Presentation states can be sent to a PACS (Provided the PACS supports the Presentations state SOP classes) and retrieved when required.
PresentationStates can be separated objects in a DICOM Study, but can also be such simple data as the Windowing setting of the image (which is part of the DICOM Image header). PyDICOM is a lowlevel library. If your PyDICOM application is showing the original pixel values and not their windowed or "presented" version, the image will look different. A DICOM viewer will most probably show the image after the presentation state is taken into consideration.
I never used PyDICOM, so I'm not sure if this is the problem, but people have reported issues here: https://github.com/pydicom/pydicom/issues/515
Please ket me know if this helped!
As Aldo pointed out, it probably is caused by different level/window (=brightness/contrast) settings. The 16 bits monochrome CT image needs to be mapped to an 8 bits monochrome image in order to be displayed, and that can be done in many ways.
I attached three 8 bit images of the same 16 bit CT slice:
- Level 1800, window 2000
- Level 1300, window 400
- Level 1000, window 100
- Badges
- Science topic
- Similar topics
- Engineering
- Electrical Engineering
- Signal Processing
- Image Processing