Skip to content
volko

Work Clinical Interoperability · proprietary

Clinical-Trial Imaging Platform

The production imaging pipeline for the Xeltis aXess FDA Breakthrough-Device trial — email intake, token uploads, five-phase DICOM verification, pixel anonymization, and PACS ingestion.

107

commits · primary author

5-phase

DICOM verification pipeline

FDA

Xeltis aXess Breakthrough Device trial

Five-stage architecture diagram of the clinical-trial imaging pipeline: study intake with tokenized upload links, per-visit DICOM upload with completeness gating, DICOM verification and QC including burned-in-PHI OCR, pixel-domain anonymization, and reconciled PACS ingestion into a PostgreSQL store.
End-to-end imaging pipeline for the trial: token-based intake, completeness-gated upload, a multi-check verification and QC pass, pixel anonymization of burned-in PHI, and reconciled PACS ingestion.

Problem

A multi-site FDA Breakthrough-Device cardiovascular trial needs imaging collected from many sites, verified, anonymized, and ingested — reliably, auditably, and without a coordinator hand-checking every study.

Architecture

An end-to-end pipeline from the trial’s email system to the PACS.

  • Intake: IMAP monitoring parses trial emails (study, site, subject, visit, assessor) and de-duplicates on message ID.
  • Token uploads: UUID tokens drive per-visit upload links to a web portal with a task checklist; chunked uploads assemble atomically.
  • Five-phase verification: header validation against the trial metadata and a SOP-class whitelist; transfer-syntax decompression; pixel anonymization via a MIRC-CTP pipeline for burned-in annotations; form detection, overlay stripping, and metadata relabeling; then C-STORE ingestion with registration confirmation and series/image-count checks.
  • Dashboard: a live admin view with per-task status, resend/cancel, and confirmation emails.

Results

In active production use for the Xeltis aXess™ trial — the first bioabsorbable cardiovascular conduit — as primary author across 107 commits.

Impact

Trial sites upload imaging through a guided, anonymizing, self-verifying pipeline instead of a manual back-and-forth — the boring infrastructure that makes a regulated trial run.