dicom-anonymizer

What This Skill Does

The dicom-anonymizer is a specialized clinical-grade utility designed to strip Protected Health Information (PHI) from medical imaging files. By targeting the 18 critical DICOM tags defined by HIPAA safe harbor standards, this skill ensures that patient identifiers—such as names, IDs, and birth dates—are permanently removed or pseudonymized. Unlike basic metadata strippers, this tool understands the complex hierarchical structure of DICOM files, allowing it to preserve essential study, series, and image data so that the resulting files remain fully functional for longitudinal research, teaching, or secondary analysis without compromising patient privacy.

Installation

To integrate this tool into your OpenClaw environment, use the official installation command: clawhub install openclaw/skills/skills/aipoch-ai/dicom-anonymizer Once installed, ensure your environment has access to the necessary Python dependencies for DICOM parsing. The skill is designed to work seamlessly within your existing local file system workflows.

Use Cases

• Clinical Research: Anonymizing large repositories of radiology scans before uploading them to collaborative cloud research portals.
• Educational Databases: Preparing de-identified case studies for medical students or training diagnostic AI models.
• Compliance Auditing: Rapidly scrubbing batch archives of patient data to ensure adherence to institutional HIPAA requirements.
• Data Sharing: Sharing anonymized scans with third-party vendors or collaborators while maintaining the integrity of image metadata.

Example Prompts

1. "Anonymize all DICOM files in the /data/mri-scans directory and save them to /data/clean-scans, ensuring all study relationships are preserved."
2. "Run the dicom-anonymizer on patient_chest_xray.dcm but make sure to keep the PatientAge tag as it is required for my current research project."
3. "Process the batch of scans in /imaging/new, generate a full audit log at /logs/anonymization_audit.json, and skip existing files to avoid duplicates."

Tips & Limitations

• Always Audit: While the tool is compliant with standard HIPAA practices, always verify the output using a DICOM viewer to ensure no sensitive tags were missed in non-standard vendor-specific private tag headers.
• Batch Limits: For extremely large datasets (10,000+ files), we recommend processing in smaller batches to monitor system memory and disk I/O performance.
• Pseudonymization: Use the --preserve-studies flag carefully; while it maintains data consistency, ensure the mapping logic aligns with your institution's specific data security protocols for linking research subjects.