DICOM de-identification tool overview

From Commontk
Jump to navigationJump to search
Home < CTK-Hackfest-Nov-2014 < DICOM de-identification tool overview

Specific use cases and requirements for de-identification

Requirements

  • Client side processing
  • Offline processing
  • Platform and browser independent
  • Provided as commandline tool?
  • In-application processing should be possible?
    • e.g. in MITK and Slicer?
  • Specify conditions like if tag value is equal to XYZ then remove all private tags except of <tag list>

Some issues, which anonymization tools often have, found on the Plastimatch homepage, also providing a comparison:

  • Some date fields cannot be changed (duh)
  • It is not possible to preserve relative dates (for example, we would like to set StudyDate to Jan 1, 2000, but preserve the fact that StructureSetDate occurred 35 days later)
  • Private tags are always deleted (on our GE scanner, these contain important acquisition details)
  • Comments, diagnosis descriptions, or other fields cannot be reset (may contain protected information)
  • It is not possible to modify strings (for example, the physician might type “smith-final” into the StructureSetName field, and we would like to delete “smith” but preserve “final” )
  • UIDs are not changed, or relationships between UIDs are not preserved (duh)

Usecases

Available tools for DICOM de-identification

Toolname Language Available as cmd line? Features URL
Clinical Trial Processor Java http://mircwiki.rsna.org/index.php?title=CTP-The_RSNA_Clinical_Trial_Processor
DICOM Cleaner Java http://www.dclunie.com/pixelmed/software/webstart/DicomCleanerUsage.html
DICOM Confidential Java http://sourceforge.net/projects/privacyguard/
DICOM Browser Java Yes Script based processing: Assignment / Deletion / Constraint and condition based / Built-in functions / Generators / user defined variables http://nrg.wustl.edu/software/dicom-browser/
GDCM Anon C++ Yes http://gdcm.sourceforge.net/html/gdcmanon.html
DCMTK C++
DICOM anon Python Yes https://github.com/cbmi/dicom-anon