dicom_redact - save redaction rectangles to sidecar file

[howff]

See also https://git.ecdf.ed.ac.uk/SMI/service/-/issues/154

As part of the extraction pipeline researchers will require access to the list of rectangles that were redacted.

• add to a DICOM tag?
• add separate file to sit alongside the DICOM?
• report all rectangles in a single report file or database?

Possible options:

• add to Image Comments tag (0020,4000) but this may already be populated, and it's non-standard
• add a Private tag, but all private tags are removed by CTP so adding a new one would be confusing
• add to a Modified Attributes Sequence tag, but this isn't a perfect match to the intention of the tag
• DICOM-RT ROI, but this is a separate file
• Presentation State DICOM file, but this is a separate file
• Add an overlay with type "ROI" to be used as a mask, but this may be difficult for the user to use
• Save an additional file alongside the DICOM file, containing rectangles in JSON format
• Pass the rectangle onto the next stage of the extraction pipeline in a rabbitmq message

[rkm]

My preference would be to avoid adding anything directly to the DICOM metadata, for the reasons you've highlighted. Easiest to get us moving would likely be a single file containing all the rectangles for the cohort, but longer term we could look at generating sidecar files, either JSON or DIOCM (SR or segmentation as suggested by Clunie) which can be provided alongside the de-identified images.

[howff]

This has been implemented in https://github.com/SMI/dicompixelanon/commit/882590bde46325f20758cc180f76d28129605ed6

dicom_pixel_anon.sh writes rectangles.csv containing columns filename,left,top,right,bottom,frame,overlay