[ITK-users] gdcmIO reader has known bugs?

[Dr Nick Patterson]

Hi,
   I have been using the gdcmIO image reader process dicom images. I wondered if anyone else regularly uses this and wanted to know if anyone else has found that it is exceptionally buggy? I have had a number of issues with ITK examples, and when I have started to adapt code to my own uses I have better understood where my programs are failing and why the outputs don’t look as they should. 

   It seems that the GDCM integration seems to struggle badly in correctly reading DICOM tags. For example, ImagePatientPosition tags are read correctly in some instances but in others the Z component it changed to 1. This is the same for the Origin tag which is changed to [0,0,0]. Having a look around some forums, there have been a number of posts regarding issues such as this over the years. Also, even in cases where the DICOM file header is correctly read in, when writing back out to another DICOM file the tags are changed to incorrect values. For example, the read in value of Rescale Intercept is -1024, but is changed to 0 when the header is copied to a new DICOM file.


I find this to be strange behaviour but was interested to know of other peoples experiences using GDCM alongside ITK.

Nick.