ITI Planning Committee 2025 Meetings
De-Identification Profile Proposals | Thursday, January 9th @ 7pm CT/8pm ET
Present: Essien Ge, John Moehrke, Soyean, Kyoohyung Han, Matt Blackmon, Chris Melo, Martin Rosner, Spencer LeGesse, Alan Zhang, Lisson Zhang Discussion points:
Comparison of Two De-Identification Profiles
De-Identification Handbook Update (Focus on PETs and Genomic Data)
Focus
- Targets updating the existing IHE De-Identification Handbook to address collaborative data linking challenges.
- Emphasizes **Privacy Enhancing Technologies (PETs)**, particularly for genomic data and cross-organizational use cases.
Problem Addressed
- Lack of common best practices for emerging healthcare data modalities (e.g., genomic data) and collaborative analysis workflows.
- Security and privacy risks in data sharing due to the absence of standardized threat models and cryptographic methods.
Key Use Cases
- Cryptographic tokenization for private data matching without revealing personal information.
- Secure genomic data sharing using cryptographic PETs.
- Workflow combining FHIR and Multi-Party Computation (MPC) modules for increased interoperability.
Proposed Updates
- Updates to the handbook incorporating PETs and genomic use cases.
- Examples of cryptographic workflows and open-source reference implementations.
Two/Multi-Stages De-Identification Process
Focus
- Introduces a two-stage de-identification process to address challenges with overly strict de-identification profiles.
- Focuses on **incremental de-identification**, allowing for regulatory-required attributes (e.g., patient age, weight) to be retained initially and removed later.
Problem Addressed
- Strict de-identification profiles (e.g., DICOM Basic Profile) remove necessary demographic data, making datasets unusable for AI development and regulatory compliance.
- Rework and delays caused by inadequate initial de-identification processes.
Key Use Cases
- Minimal de-identification performed at the modality/PACS level.
- Additional de-identification by experts to meet specific data-sharing requirements for secondary use (e.g., AI development).
Proposed Updates
- Introduces a stepwise approach to de-identification, allowing for staged compliance.
- Addresses usability gaps by ensuring datasets meet the needs of regulatory and secondary users.
Discussion Points
- De-identification handbook can be updated with both use cases
- Each use case may result in a profile
- e.g., last de identification handbook resulted in this whitepaper
- Philips presented detailed use case for two/multi stage process
- Soyean provided details on genomic data/linking use case
Actions
- Each team will update respective proposals based on discussion
- Each team will provide more background on each proposal
- Follow-up meeting scheduled Feb 13th after Chinese New Year
- @John to create GitHub repo for working documents
- Those not already in itiplan and ititech mail lists to follow-up with Sara Bell