nitrc:De-Identification Agreement
From NITRC Wiki
Contents |
Agreement
The Privacy Act of 1974, 5 U.S.C. ยง 552a, protects records that can be retrieved by personal identifiers. Examples of identifying pieces of information (taken from the Center for Information Technology, National Institutes of Health) can include:
- Personal characteristics (such as height, weight, gender, sexual orientation, date of birth, age, hair color, eye color, race, ethnicity, scars, tattoos, gang affiliation, religious affiliation, place of birth, mother's maiden name, distinguishing features and biometrics information, such as fingerprints, DNA and retinal scans).
- A unique set of numbers or characters assigned to a specific individual, including name, address, phone number, SSN, e-mail address, driver's license number, financial account or credit card number, and Automated Integrated Fingerprint Identification System (AIFIS) identifier, booking, or detention system number.
- Descriptions of events or points in time (for example, information in documents such as police reports, arrest reports and medical records).
- Descriptions of locations or places, including geographic information system (GIS) locations, electronic bracelet monitoring information, etc.
Data uploaded to NITRC's Image Repository, NITRC-IR, shall not contain personally identifiable information. Please check with your institutional review board (IRB) for policies related to de-identifying your data. By checking the box that appears when uploading data to NITRC, you have agreed to provide only de-identified data to NITRC's federation data repository and certified that, to the best of your knowledge, any personally identifiable information has been removed from your data in accordance with applicable IRB policies.
De-Identification Tools
Below is general information about data de-identification as well as existing solutions. Please help the NITRC Community by updating this page with other information or solutions regarding de-identification.
General Information
Please note that the de-identification solutions on this page are intended as a starting point until you can verify all problematic fields are cleared, removed, or pseudonymized. NITRC makes no guarantee that at all potentially identifying information is removed by using any of these solutions. Best practices include testing these solutions with your data and then using multiple tools to manually verify that all personally identifiable information (PII) has been removed. It is recommended that the output is reviewed using at least a full metadata display tool like dcmdump and a string dump utility such as UNIX's strings.
Background information on data de-identification in the United States can be found on the NIH's HIPAA Privacy Rule site and in the NIH Guide for Identifying Sensitive Information.
DICOM Data
General Information
The DICOM standard addresses de-identification in Annex E (Attribute Confidentiality Profiles) of Part 15 of the standard. DICOM supplements and corrections apply to this section and should also be consulted.
The National Cancer Institute's DICOM de-identification page is a more accessible introduction to the subject.
Tools
Many tools are available for de-identifying DICOM data; this is a partial list of those tools. See also David Clunie's list of DICOM Anonymizers.
- GDCMANON is a command-line tool that implements de-identification of DICOM files according to the DICOM Supplement 55: Attribute Level Confidentiality DICOM specification.
- CCB DICOM Anonymizer is a pipeline tool which may be harder to access in a batch format.
- DICOMBrowser is an application for inspecting and modifying many DICOM files at a time.
NIfTI-1 Data
General Information
While most (but not all) of the basic NIfTI-1 header is immune to identifying information, NIfTI-1 allows for header extensions that may contain identifying information. Often a simple conversion to NIfTI-1 will result in de-identified data, but proceed with care.
Tools
- LONI De-Identification Debablet has a pipeline Java application for removing patient-identifying information (e.g., patient name and id) from medical image files
- The standard NIfTI libraries allow for direct access to NIfTI files.
