Anonymising JPEG or other embedded tags

[jas88]

Since the DICOM file usually contains an embedded JPEG or TIFF file as the "pixel data", we should probably extract and examine the metadata within that as we already do for DICOM tags and strip/redact as appropriate? It won't be directly patient-specific, but things like timestamps could be an issue we'd care about, particularly if we're rounding off the timestamps in the DICOM tags.

[rkm]

Great idea.

It looks like we don't extract this info at the moment unfortunately, since fo-dicom stores this inFileMetaInfo rather than the Dataset object. So we can implement this and it will be stored for any new files we scan, but going back and retroactively adding it to the NPACS data will take some effort...

[jas88]

Maybe something to check for during the extraction/redaction pipeline rather than extracting beforehand - I don't think there's anything we'd want to index/promote for cohort building, just check we aren't leaking anything unwanted through the anon pipeline?