Clinical Data Pipelines
Config-driven ETL for imaging and clinical data, built to be audited.
Learn more →PHI hides in more places than the header
DICOM header scrubbing is table stakes. The PHI that actually gets organizations in trouble is burned into pixel data — ultrasound and secondary-capture screen text, C-arm overlays, dose reports, scanned requisitions folded into the study. A de-identification process that only touches header tags will pass a checklist and still leak patient names.
What we do
- Header-level de-identification — configurable tag scrubbing aligned to your data-use agreement, not a one-size-fits-all profile.
- Pixel-level de-identification — OCR-based census of burned-in text across modalities, with visual review of flagged frames before release.
- Audit trails — a defensible record of what was found, what was removed, and who reviewed it, built for compliance and research IRB requirements.
- Pipeline integration — de-identification wired directly into your ingestion or research pipeline, not a manual side process someone forgets to run.
We built OCR-driven pixel census tooling specifically because manual visual review alone doesn't scale past a few hundred studies — and spot-checking isn't a defensible process for PHI.
Who this is for
Research teams building imaging datasets, AI vendors sourcing training data, and health systems sharing data with external partners — anywhere de-identified imaging needs to hold up to scrutiny, not just pass a quick glance.