DICOM Metadata & PHI Anonymization

Extract, inspect, and anonymize metadata from DICOM (Digital Imaging and Communications in Medicine) medical image files. This skill provides AI agents with the ability to parse DICOM headers for study and patient information, batch-extract metadata fields across imaging series, identify and redact protected health information (PHI) embedded in DICOM tags for de-identification, and prepare imaging datasets for research use in compliance with HIPAA Safe Harbor and Expert Determination methods -- powered by the pydicom library.

Quick Install

npx skills add Open-Medica/open-medical-skills --skill dicom-metadata-extractor

What It Does

• DICOM header parsing to extract all standard metadata fields including Patient ID, Study Date, Modality, Body Part, Series Description, Institution Name, Referring Physician, pixel spacing, slice thickness, and hundreds of additional DICOM tags organized by group
• Batch metadata extraction across entire imaging studies or series, generating structured CSV or JSON reports of metadata fields for all files in a directory, enabling systematic data audits
• PHI identification and removal following HIPAA Safe Harbor de-identification (18 identifiers), automatically detecting and redacting patient name, date of birth, medical record number, accession number, institution name, referring physician, and other PHI-containing DICOM tags
• Pixel data PHI detection for burned-in annotations, identifying text overlays on image pixels that may contain patient identifying information (e.g., patient name rendered in image corners by the modality)
• Research dataset preparation with configurable anonymization profiles, date shifting (preserving temporal relationships between studies), UID remapping (maintaining series/study relationships while replacing identifiers), and export in standard DICOM or converted formats

Clinical Use Cases