File Processing | Batch Processing |
---|---|
deidentify is a tool to remove personal identifiers from free-text medical record data. Detected identifiers are replaced by randomly generated substitutes. Consistency of the data is preserved as the same name, phone number or location will always be mapped to the same replacement.
- Facilities to remove all relevant identifiers of individuals from medical record information to comply with the HIPAA "Safe Harbor" rule
- Single file and batch processing
- Customizable options
- Persistent data store ensures consistency of mappings and allows re-identification of the de-identified data
The deidentify tool uses several open-source projects.
The desktop application was created with nw.js, formerly called node-webkit, and is entirely written in JavaScript.
Our de-identification procedure combines hand-crafted regular expressions with the named entity recognizer (NER) developed by the Stanford Natural Language Processing Group, which provides a Conditional Random Field (CRF) model for detecting the three classes PERSON, ORGANIZATION, LOCATION.
Reference:
Finkel, J. R., Grenager, T., & Manning, C. (2005). Incorporating non-local information into information extraction systems by gibbs sampling. In Acl, (1995), 363 – 370. doi:10.3115/1219840.1219885
To generate random replacements for detected identifiers, the chance.js library is used. NeDB is used as a data store, keeping track of the mappings from original identifiers to replacements.
Other used libraries include
This software uses the Stanford NER tool, which requires Java 1.8 or later.
Installers for Windows, MacOS and Linux can be downloaded from the releases page.
Clone the repository via the following command:
git clone --recursive https://github.com/Planeshifter/deidentify.git
Change into the newly created directory, install npm dependencies and run the init
script:
npm install
npm run init
Start the program by executing the following command from the project directory:
npm start
This project is licensed under the GNU General Public License v2.0.
Copyright © 2015-2018. Philipp Burckhardt.