MIRC-CTP IRT Anonymization and Filter scripts
DISCLAIMER: These anonymization scripts are only provided for testing the MIRC-CTP DICOM file output with your application. They are not intended to be used in a clinical or research setting, and should be considered incomplete test samples. DICOM files filtered through this program and associated scripts are not guaranteed to be free of PHI.
This project contains baseline MIRC-CTP de-identification and filtering scripts used within Stanford IRT-RIT for anonymizing DICOM studies at scale. Use these scripts to verify that the Stanford IRT-RIT de-identification pipeline produces output acceptable for your study.
Note
These scripts are oriented towards removing PHI and images that are not useful for machine learning. Image types that are "DERIVED" or "SECONDARY" are excluded, as they are generally not useful for machine learning and are far more likely to contain pixel-PHI. If you modify these scripts to include "SECONDARY" or "DERIVED" it is very likely the pixel scrubbing scripts will pass-through images that still contain pixel-PHI.
Included in this project is the MIRC-CTP command-line DicomAnonymizerTool which allows de-identification of DICOM studies without installing the entire MIRC-CTP application. The Stanford IRT-RIT anonymization pipeline uses this same library.
DICOM anonymization scripts
stanford-anonymizer.script: This file specifies which DICOM tags should be modified or removed.stanford-filter.script: This file specifies which DICOM instances should be removed. Currently includes image types known to have pixel data with PHI, for example secondary derived screens (screenshots).stanford-scrubber.script: MIRC-CTP standard pixel scrubbing definitions with additional rules added by Stanford.
The anonymization scripts are based off the DICOM-PS3.15E-Basic profile with additional rules for tags known to contain PHI. All vendor-specific (eg. odd-numbered) tags are also removed.
A DICOM tag reference can be found here.
Requirements
Since we run this tool on MacOS, it requires docker (see below)
Installation
First ensure you have the Oracle JDK v.8 installed.
Create a clone of this repository on your workstation:
git clone --recurse-submodules https://github.com/susom/mirc-ctp.git
If you do not have the ant program installed, install it with HomeBrew (which will need to be installed if you haven't done so already)
$ brew install ant
Compile the included DicomAnonymizerTool by typing ant at the command prompt:
$ ant
Buildfile: /Users/jdoe/Projects/mirc-ctp/build.xml
clean:
init:
[echo] Time now 15:56:40 PST
[echo] ant.java.version = 1.8
[mkdir] Created dir: /Users/jdoe/Projects/mirc-ctp/DicomAnonymizerTool/build
...
You should now have a directory called DAT which contains the DicomAnonymizerTool. You can try running it:
$ java -jar DAT/DAT.jar
Usage: java -jar DAT {parameters}
where:
-in {input} specifies the file or directory to be anonymized
If {input} is a directory, all files in it and its subdirectories are processed.
-out {output} specifies the file or directory in which to store the anonymized file or files.
...
Now the application needs to be placed in a Docker image. To create the image:
docker build -f Dockerfile --pull -t mirc-ctp .
You can now place some test DICOM studies in the directory DICOM and run the shell script which will anonymize the studies (all to the same anonymous MRN and Accession Number) and place them in DICOM-ANON
$ ./anonymize.sh
----
Thread: pool-1-thread-1: Anonymizing DICOM/1.2.840.4267.32.293501795892579834759834759834759834
Anonymized file: DICOM-ANON/1.2.840.4267.32.10027221686667529588514012002002498656
----
Thread: pool-1-thread-2: Anonymizing DICOM/1.2.840.4267.32.093248509348509384509384509834059840
Anonymized file: DICOM-ANON/1.2.840.4267.32.10134745174550989356450666756661275833
----
Elapsed time: 0.634
You can now open the DICOM files in DICOM-ANON to make sure they work with your intended application.
Pixel filtering and MacOS
In order to read DICOM encoded with the JPEG Lossless syntax, you need to have the Java Advanced Imaging ImageIO libraries. Unfortunately, these are not available for Mac. To get around this limitation, this application executes within Docker. A Dockerfile is included in this distribution.