The DICOM files will be examined by an OCR algorithm to find text burned into the pixels of any of its image or overlay frames. All text found will be redacted by replacing it with a black rectangle.
The location of all text can be kept in a database or exported to a CSV file in case it is needed later.
The program calls dicom_ocr.py to run OCR on the image frames and overlay
frames in one or more DICOM files, saving the results in a database, and then runs
dicom_redact.py to actually redact the image pixels, saving the resulting images in
new DICOM files.
dicom_pixel_anon.py [-D db dir] -o output input...
-D database directory, optional, to persist rectangle locations
-o output directory, optional, default same as input, or current dir
-r relative directory, optional, see `dicom_redact.py` for explanation
input is one or more DICOM filenames.
A record of the redacted rectangles is saved in a CSV file called
rectangles.csv in the output directory. The columns will be
filename,left,top,right,bottom,frame,overlay. This file is only written
after all DICOM files have been anonymised.
When used as part of an anonymisation pipeline you should ensure that the database directory is not part of the extracted data hierarchy because it will contain OCR text extracted from the DICOM files and thus possibly PII. In fact, after anonymising a set of DICOM files the database directory can be removed if you do not need it for future redactions.
This program requires dicom_ocr.py, dicom_redact.py - see their documents
for more details on their requirements.
It optionally uses the sqlite3 command line utility to extract from the
database into a CSV file.
See the dicom_ocr.py document.
See the dicom_redact.py document.