18. Exercise 3: Anonymization

I hope you enjoyed this little exercise and have a better grasp of what is involved in anonymization of DICOM datasets. You can find the solution to this exercise here. As you inspect the solution, take note of how HIPAA gives you a good start for what is considered a PHI and what isn’t, but really there is much more that you should be aware of, even when anonymizing just the DICOM tags.

While it is fun to get hands-on and look at those individual tags and algorithms yourself, there are libraries out there that do a good job of de-identifying DICOM and it’s something you should consider when setting up de-id pipelines.


An example of a good library is the deid module of pydicom that cleans both metadata and image pixels: https://pydicom.github.io/deid/


Another such library is the skull stripping functionality of the popular neuroimaging FreeSurfer package. This package can remove pixels containing facial features while leaving brains intact.