INDUSTRY FIELD GUIDE
Diagnostic imaging worklist and export verification data for AI
An export can omit an identifying metadata field while text is still visible in the image pixels.
NAICS 621512 · Diagnostic Imaging Centers
An export can omit an identifying metadata field while text is still visible in the image pixels. DICOM’s confidentiality profiles describe separate handling for attributes and pixel data. An imaging center can build a practice inspection using original phantom or synthetic images. DICOM confidentiality profiles
Generate two abstract images with mock study identifiers. Make one header match the worklist while a visible overlay carries a dummy name. Swap a second export’s study identifier. Ask which files need review and have the evaluator identify the exact field or region. Keep image interpretation outside the task.
Use generated images or nonpatient training phantoms. Exclude patient scans, live worklists and device records. Have an authorized specialist verify any future real data route.
Describe a synthetic imaging export review
Original Fiund exercise
Removing a header field does not remove a visible text overlay
Two independently generated abstract images are represented by original metadata and region annotations. They contain only dummy names and invalid study aliases, with no patient scan or device export.
| Record | Source | What it records |
|---|---|---|
| IX1 | Mock worklist | Exercise E lists study DEMO-A for image A and DEMO-B for image B. |
| IX2 | Image A review | Header study DEMO-A matches. Header name field absent, but original pixel-overlay annotation identifies dummy text 'Practice Person' in the upper-right region. |
| IX3 | Image B review | Header study says DEMO-A, conflicting with E's DEMO-B reference. No lineage record explaining the swap; no visible-name finding supplied for B. |
Question for the model
Which exact review findings apply to A and B, and can missing header text establish that A's image pixels contain no identifying overlay?
Show the reference answer and checks
IX2 supports A's study-header match and separately flags its upper-right dummy-name overlay. An absent header name does not remove text already drawn into pixels. IX3 supports a study-reference conflict for B, not a conclusion about B's pixel content. Return those file-specific findings and regions for specialist review rather than treating one cleared attribute as complete de-identification or silently correcting an unexplained study swap.
- Assuming removal of metadata removes burned-in image text.
- Overwriting a study identifier without resolving its mismatch or copying findings across files.
Fields to preserve
- Exercise and image/study reference
- Defines the expected linkage.
- Header attribute and observed value
- Supports metadata review.
- Pixel region and review state
- Separates visible content from header fields.
Test another case
Remove A's overlay annotation but provide no pixel review. Require its visible-text state to become unknown rather than presume clear pixels.
Use original synthetic images or approved nonpatient phantoms. Exclude scans, live worklists and device records; no clinical interpretation or complete privacy determination is performed.
Discuss a contribution
Describe the sources you control, the task they could support and what has already been checked. Use the existing-data assessment for an archive or the collection brief for a new recording project. The contributor agreement defines Fiund’s licensing authority and compensation. A guide or fictional exercise does not establish an accepted collection or a sale.