Saltar al contenido principal

DICOM Review Queue

Category: Healthcare Standards
Node id: dicom-review-queue
Maturity: production
Risk tier: C

Opens a human review request for every de-identified instance the detector declined to clear, so escalated images are recorded instead of lost.

What it does​

  • Opens one review per instance the detector declined to clear, carrying the reasons and the detection geometry
  • Idempotent per SOP Instance UID: a retried study does not file the same image twice
  • Carries references, never pixels — the image stays in the source archive
  • Queues a de-identifier failure too, labelled apart from a deliberate escalation

Use cases​

  • Routing images with burned-in PHI the detector could not clear to a human reviewer
  • Keeping a research export auditable: every image withheld from publication leaves a record
  • Telling a de-identifier failure apart from a clinical escalation, which look identical downstream

Runtime​

  • Performance tier: fast

Input ports​

NameTypeRequiredDescription
instancesjsonYesThe dicom-deid result (array mode): result.instances[]. Emitted instances are ignored; escalated and errored ones are queued.

Output ports​

NameTypeRequiredDescription
resultjsonNoWhat the run did: reviews opened, already-queued instances skipped, and what could not be queued and why. Whether the STUDY is waiting for a person is reported by dicom-deid, which is the node that knows what it emitted — and which lands a batch earlier, where a marking step still runs when publication fails.
reviewsjsonNoThe review requests opened by this run, each linked to the instance it is about.
countnumberNoHow many reviews this run opened.

Settings​

Basic​

SettingNameTypeRequiredDefaultAllowed valuesDescription
AccessionaccessionstringNoThe episode the run is driven by, recorded on each review so a reviewer can find the study.
Concentrator Base URLconcentrator_base_urlstringNoUsed to build the WADO-RS reference to the ORIGINAL instance. No pixels are copied; the reviewer fetches it through an authorised path.
PrioritypriorityenumNohighlow, medium, high, urgent, criticalPriority of the reviews this run opens. High by default: an unreviewed escalation is an image withheld from the study it belongs to.
Required Rolerequired_rolestringNoimaging_reviewerRole a reviewer must hold. These rows carry original identifiers.

Advanced​

SettingNameTypeRequiredDefaultAllowed valuesDescription
Workflow IDworkflow_idstringNodicom-deid-reviewGroups these reviews in the inbox and is recorded on each request, so imaging work can be told from other review traffic.
Escalation Hoursescalation_hoursnumberNo24Hours before an untouched review escalates. The image is already withheld, so the cost of waiting is a gap in the study, not an exposure.

Errors​

CodeDescription
REVIEW_STOREExecuted without a human-review store (no platform database)
REVIEW_INPUTThe instances input is missing or is not a list

Runtime notes​

  • Requires the platform database: registered without a store and wired at server boot; reports a clear error when it is missing rather than silently leaving escalated images unreviewed.
  • Idempotent per SOP Instance UID, so a study that retries does not file the same image again.
  • Carries REFERENCES and findings, never pixels. The row holds original PACS identifiers, so it is marked contains_phi and belongs under the same restriction as the crosswalk.
  • An instance the de-identifier ERRORED on is queued too, labelled apart from a deliberate escalation: both need a human, and a crash must not pass for a clinical judgement.

See Build any workflow for how to put nodes together and Ports, types and data flow for how port types connect.

¿Algo no está claro? Envíe sus comentarios sobre la documentación por correo electrónico..