Python package to anonymize DICOM files. The anonymization answer to the standard . More information about dicom fields for anonymization can be found here.
The default behaviour of this package is to anonymize DICOM fields referenced in the 2023e DICOM standard. These fields are referenced in dicomfields.
Another standard can be selected, see Change the DICOM anonymization standard.
Dicom fields are separated into different groups. Each groups will be anonymized in a different way.
Group | Action | Action definition |
---|---|---|
D_TAGS | replace | Replace with a non-zero length value that may be a dummy value and consistent with the VR** |
Z_TAGS | empty | Replace with a zero length value, or a non-zero length value that may be a dummy value and consistent with the VR** |
X_TAGS | delete | Completely remove the tag |
U_TAGS | replace_UID | Replace all UID's random ones. Same UID will have the same replaced value |
Z_D_TAGS | empty_or_replace | Replace with a non-zero length value that may be a dummy value and consistent with the VR** |
X_Z_TAGS | delete_or_empty | Replace with a zero length value, or a non-zero length value that may be a dummy value and consistent with the VR** |
X_D_TAGS | delete_or_replace | Replace with a non-zero length value that may be a dummy value and consistent with the VR** |
X_Z_D_TAGS | delete_or_empty_or_replace | Replace with a non-zero length value that may be a dummy value and consistent with the VR** |
X_Z_U_STAR_TAGS | delete_or_empty_or_replace_UID | If it's a UID, then all numbers are randomly replaced. Else, replace with a zero length value, or a non-zero length value that may be a dummy value and consistent with the VR** |
ALL_TAGS | Contains all previous defined tags |
Installation can be done via pip pip install dicom-anonymizer
or conda conda install -c conda-forge dicom-anonymizer
.
To get started with local development, follow these steps:
-
Create a Virtual Environment:
- On Windows:
virtualenv env .\env\Scripts\activate.bat
- On MacOS/Linux:
python -m venv env source env/bin/activate
- On Windows:
-
Install Dependencies:
- Install an editable version of the package and the development requirements:
pip install -e .[dev]
- Install an editable version of the package and the development requirements:
-
Set Up Pre-Commit Hooks:
- Install the pre-commit hooks to ensure code quality:
pre-commit install
- Install the pre-commit hooks to ensure code quality:
To run the unit tests, use the following command:
pytest
These instructions rely on wheel build-package format. Install it if you have not done it already using:
pip install wheel
The sources files can be packaged by using:
python ./setup.py bdist_wheel
This command will generate a wheel package in dist
folder which can be then installed as a python package using
pip install ./dist/dicom_anonymizer-1.0.13-1-py2.py3-none-any.whl
On Windows, if you see a warning message
'./dist/dicom_anonymizer-1.0.13-1-py2.py3-none-any.whl' looks like a filename, but the file does not exist
,
this could be due to pip not being able to handle relative path (See issue pypa/pip#10808). As a work-around, change directory to dist
and then install it using
pip install dicom_anonymizer-1.0.13-1-py2.py3-none-any.whl
Installing this package will also install an executable named dicom-anonymizer
. In order to use it, please refer to the next section.
This package allows to anonymize a selection of DICOM field (defined or overridden). The way on how the DICOM fields are anonymized can also be overridden.
- [required] InputPath = Full path to a single DICOM image or to a folder which contains dicom files
- [required] OutputPath = Full path to the anonymized DICOM image or to a folder. This folder has to exist.
- [optional] ActionName = Defined an action name that will be applied to the DICOM tag.
- [optional] Dictionary = Path to a JSON file which defines actions that will be applied on specific dicom tags (see below)
You can use the default anonymization behaviour describe above.
dicom-anonymizer Input Output
Default behavior of the dicom anonymizer is to delete private tags. But you can bypass it:
- Solution 1: Use regexp to define which private tag you want to keep/update (cf custom rules)
- Solution 2: Use dicom-anonymizer.exe option to keep all private tags :
--keepPrivateTags
You can manually add new rules in order to have different behaviors with certain tags. This will allow you to override default rules:
Executable:
dicom-anonymizer InputFilePath OutputFilePath -t '(0x0001, 0x0001)' ActionName -t '(0x0001, 0x0005)' ActionName2
This will apply the ActionName
to the tag '(0x0001, 0x0001)'
and ActionName2
to '(0x0001, 0x0005)'
Note: ActionName has to be defined in actions list
Example 1: The default behavior of the patient's ID is to be replaced by an empty or null value. If you want to keep this value, then you'll have to run :
python anonymizer.py InputFilePath OutputFilePath -t '(0x0010, 0x0020)' keep
This command will override the default behavior executed on this tag and the patient's ID will be kept.
Example 2: We just want to change the study date from 20080701 to 20080000, then we'll use the regexp
python anonymizer.py InputFilePath OutputFilePath -t '(0x0008, 0x0020)' 'regexp' '0701$' '0000'
Example 3: Change the tag value with an arbitrary value
python anonymizer.py InputFilePath OutputFilePath -t '(0x0010, 0x0010)' 'replace_with_value' 'new_value'
DICOMDIR anonymization is not specified. It is therefore discouraged and it is recommended to regenerate new DICOMDIR files after anonymizing the original DICOM files.
DICOMDIR files can have a (0x0004, 0x1220) Directory Record Sequence
tag that can contain patient information.
However, this tag is not part of the standard tag to anonymize set. If you still want dicom-anonymizer to anonymize it, you have to instruct it explicitly:
python anonymizer.py InputFilePath OutputFilePath -t '(0x0004, 0x1220)' replace
Instead of having a big command line with several new actions, you can create your own dictionary by creating a json file dictionary.json
:
{
"(0x0002, 0x0002)": "ActionName",
"(0x0003, 0x0003)": "ActionName",
"(0x0004, 0x0004)": "ActionName",
"(0x0005, 0x0005)": "ActionName"
}
Same as before, the ActionName
has to be defined in the actions list.
dicom-anonymizer InputFilePath OutputFilePath --dictionary dictionary.json
If you want to use the regexp action in a dictionary:
{
"(0x0002, 0x0002)": "ActionName",
"(0x0008, 0x0020)": {
"action": "regexp",
"find": "0701$",
"replace": "0000"
}
}
Here is a small example which keeps all metadata but updates the series description by adding a suffix passed as a parameter.
import argparse
from dicomanonymizer import ALL_TAGS, anonymize, keep
def main():
parser = argparse.ArgumentParser(add_help=True)
parser.add_argument(
"input",
help="Path to the input dicom file or input directory which contains dicom files",
)
parser.add_argument(
"output",
help="Path to the output dicom file or output directory which will contains dicom files",
)
args = parser.parse_args()
deletePrivateTags = False
input_dicom_path = args.input
output_dicom_path = args.output
extra_anonymization_rules = {}
# Per https://www.hhs.gov/hipaa/for-professionals/privacy/special-topics/de-identification/index.html
# it is all right to retain only the year part of the birth date for
# de-identification purposes.
def set_date_to_year(dataset, tag):
element = dataset.get(tag)
if element is not None:
element.value = f"{element.value[:4]}0101" # YYYYMMDD format
# ALL_TAGS variable is defined on file dicomfields.py
# the 'keep' method is already defined into the dicom-anonymizer
# It will overrides the default behaviour
for i in ALL_TAGS:
extra_anonymization_rules[i] = keep
extra_anonymization_rules[(0x0010, 0x0030)] = set_date_to_year # Patient's Birth Date
# Launch the anonymization
anonymize(
input_dicom_path,
output_dicom_path,
extra_anonymization_rules,
delete_private_tags=False,
)
if __name__ == "__main__":
main()
See the full application in the examples
folder.
In your own file, you'll have to define:
- Your custom functions. Be careful, your functions always have in inputs a dataset and a tag
- A dictionary which map your functions to a tag
You can also anonymize dicom fields in-place for pydicom's DataSet using anonymize_dataset
. See this example:
import pydicom
from dicomanonymizer import anonymize_dataset
def main():
# Create a list of tags object that should contains id, type and value
fields = [
{ # Replaced by Anonymized
'id': (0x0040, 0xA123),
'type': 'LO',
'value': 'Annie de la Fontaine',
},
{ # Replaced with empty value
'id': (0x0008, 0x0050),
'type': 'TM',
'value': 'bar',
},
{ # Deleted
'id': (0x0018, 0x4000),
'type': 'VR',
'value': 'foo',
}
]
# Create a readable dataset for pydicom
data = pydicom.Dataset()
# Add each field into the dataset
for field in fields:
data.add_new(field['id'], field['type'], field['value'])
anonymize_dataset(data)
if __name__ == "__main__":
main()
See the full application in the examples
folder.
For more information about the pydicom's Dataset, please refer here.
You can also add extra_anonymization_rules
as above:
anonymize_dataset(data_ds, extra_anonymization_rules, delete_private_tags=True)
Action | Action definition |
---|---|
empty | Replace with a zero length value, or a non-zero length value that may be a dummy value and consistent with the VR** |
delete | Completely remove the tag |
keep | Do nothing on the tag |
replace_UID | Replace all UID's number with a random one in order to keep consistent. Same UID will have the same replaced value |
empty_or_replace | Replace with a non-zero length value that may be a dummy value and consistent with the VR** |
delete_or_empty | Replace with a zero length value, or a non-zero length value that may be a dummy value and consistent with the VR** |
delete_or_replace | Replace with a non-zero length value that may be a dummy value and consistent with the VR** |
deleteOrEmptyOrReplace | Replace with a non-zero length value that may be a dummy value and consistent with the VR** |
delete_or_empty_or_replace_UID | If it's a UID, then all numbers are randomly replaced. Else, replace with a zero length value, or a non-zero length value that may be a dummy value and consistent with the VR** |
regexp | Find a value in the tag using a regexp and replace it with an arbitrary value. See the examples in this file to learn how to use. |
replace_with_value | Replace the tag value with an arbitrary value. See the examples in this file to learn how to use. |
** VR: Value Representation
Work originally done by Edern Haumont
You can customize the DICOM standard that will be used to anonymize the dataset by giving an argument base_rules_gen
to the function anonymize_dicom_file
or anonymize_dataset
.
The value should be a function returning a dict of anonymization rules. Use the function initialize_actions
to create such dict from a anonymization database from the folder dicomanonymizer/dicom_anonymization_databases
.
Example:
from dicomanonymizer.simpledicomanonymizer import anonymize_dataset, initialize_actions
anonymize_dataset(
dataset, base_rules_gen=lambda: initialize_actions("dicomfields_2024b")
)