XyDromatics VNA Migration

The role matrix — including the shared admin / operator / viewer triad and any custom roles — is managed on a dedicated page. Changes persist in role_matrix.json inside the data directory.

Loading…

Select a .lic file issued by the Synthology License Generator. The file is staged, decrypted, and — on successful parse — replaces the active license. Feature gating is re-evaluated on the next request; no restart is required.

Gates the POST /api/batch_sources/webhook seed endpoint that external systems hit to feed migration batches. When the toggle is off the endpoint is anonymous (back-compat default). API keys + optional HMAC + optional IP allowlist cover the three usual second-factors for inbound webhooks. The curl example you hand to integrators lives in External Batch Sources → REST/Webhook.

Sets the per-host identity that the in-process SupportAgent uses to connect to SynthGateway. Stored in …. Changes require a SupportAgent service restart on this host.

Current engine

—

Configured

—

Changing the archive catalog DB engine or connection string requires a service restart. The SCP will refuse to start until the catalog is reachable.

…

…

…

Read-only view filtered to nodes whose role is SourceOnly or Both. Phase 2H.2 adds the retrieval-job orchestration that consumes these source AEs for batch C-FIND / C-MOVE retrieval from legacy PACS.

POST a BatchSeed JSON body to the endpoint below. The dispatcher deduplicates by (channel, external_id) — repeat calls with the same external_id return the prior batch_id rather than re-queueing. Empty external_id always creates a new batch.

Listens on the existing MLLP port for ORM^O01 (orders) and ADT^A08 (patient updates). Each accepted message becomes a BatchSeed dispatched through the same pipeline as REST/Webhook. Per-sender mapping resolves MSH-3 SendingApp / MSH-4 SendingFacility → DICOM source-node-id; unmapped senders fall through to the catch-all (or are inbox-only when no catch-all is set).

Polls a configured directory for CSV / JSON worklist files (operators drop via cron / SFTP / batch jobs). Each parsed file becomes one BatchSeed. Successful files move to {dir}/processed/; failed files move to {dir}/failed/ with a sidecar .error.txt explaining the parse failure.

accession_number,patient_id
ACC-2026-0001,MRN001
ACC-2026-0002,MRN002

{
  "external_id": "epic-overnight-2026-05-23",
  "batch_name": "Epic overnight",
  "accession_numbers": ["ACC-2026-0001", "ACC-2026-0002"],
  "study_instance_uids": [],
  "patient_id": "",
  "force_full_repull": false
}

Polls a configured FHIR R4 base URL every PollIntervalMinutes (default 5) for new ImagingStudy + ServiceRequest resources. Uses _lastUpdated=gt{cursor} so each sweep only fetches resources changed since the prior poll. Accession is extracted from identifier.type.coding.code = "ACSN"; DICOM Study UID from identifier.system = "urn:dicom:uid".

Use when the legacy PACS has no API. Operator authors a parameterized SELECT query that returns one row per study-to-migrate; the connector executes it on a cadence, binds @since to the persisted cursor, and dispatches one BatchSeed per row.

For legacy hospital systems exposing SOAP web services (DICOMweb precursors, IHE PIX/PDQ v3, HL7v3 wrappers, vendor proprietary endpoints). Operator authors a full SOAP envelope with {since} / {limit} / {wssecurity_header} placeholders. The poller POSTs on a cadence, parses the response via XPath, and dispatches one BatchSeed per row.

Build one or more retrieval batches. Each batch picks its own source PACS, batch name, and list of accessions or study UIDs. The background MigrationRetrievalService runs C-FIND + C-MOVE for every entry in every batch.

Multi-batch tip (Phase 2J.2): use "Add another batch" to queue several batches in one submit. Batches can target different source PACS — useful for hospital-network consolidations that span multiple legacy systems.

Bulk import from delimited text or file

Paste CSV/TSV/pipe-delimited text, or upload a file (.csv / .tsv / .psv / .txt; 5 MB max). Pick the separator + column that holds the accession or study UID, then append the parsed values into the lists below.

Run a single C-FIND-STUDY query across many source PACSes in parallel. Results are merged by Study Instance UID and tagged with the sources that returned each study, so you can see at a glance which legacy systems carry a given study.

Use this for cross-network reconciliation before opening the Migration Wizard — select source PACSes, set a query, then "Send selected to Wizard" to pre-fill batch rows.

Sources to search

Query

Per-source status

All Expanded Analytics charts below respect the window + the Filter bar above (Study date / Source / Modality).

Runs first in the ingest pipeline (before Validation / MissingData / Reconciliation / TagMapping / QualityGate / Discovery). Brings non-PS3.x-compliant DICOM from direct legacy-PACS retrieval into shape so downstream engines have correct input.

Leave Enabled = off when ingesting via the Migration Engine route (data already normalized upstream). Turn on when ingesting direct from a legacy PACS via C-MOVE or file drop.

Policies

Every federated search hit upserts a (study UID, source PACS) row into this catalog. Browse the accumulated history without re-running C-FIND. Use the Tie-Breaker Rules pane to control which source the federated-search pane picks as "Candidate" when a study lives in multiple legacy systems.

Records

When a federated search finds the same Study UID at multiple source PACSes, the resolver applies these per-source-pair rules to pick a Candidate retrieval source. Rules are advisory: the federated-search pane decorates rows with the candidate but the operator still confirms the source in the wizard.

Default behavior when no rule is configured for a pair = first-found wins (unchanged from Phase 2J.3). Add a rule for any pair where you want different behavior.

Test a rule

Plug in two sources and synthetic counts; preview which one the resolver would pick under the currently-saved rule for that pair.

Auto-fired whenever a batch reaches all-terminal (every job complete / failed / cancelled). For each completed job, queries source PACS via C-FIND and compares against the local archive (series + instance counts). Studies with a delta land here for triage.

Enable auto-requeue on partial per batch to have the service spawn a fresh single-shot retrieval job for missing/partial studies automatically (max_attempts=1 to prevent loops). DCR-2026-272 (filed approved + parked) extends this with instance-level C-MOVE for the second pass.

Background service

Schema: id, name, description, tag (DICOM keyword), rule_type (format / enum / length / regex / derived / custom), enabled, on_fail (Correct / Flag / Reject / Warn / Ignore), valid_values, correction_map, pattern, format (DA / TM / AS), max_length, min_length, derivation_method (compute_age / sop_to_modality), derivation_source.

JSON dict of sample DICOM tags. Engine evaluates rules in memory; no persisted state touched.

Schema: id, name, description, tag, enabled, priority, policy (Default / Derive / Lookup / Flag / Reject / Ignore), default_value (also format-template source), derivation_method (keyword_extract / compute_age / sop_to_modality / copy_tag / format_template), derivation_source, derivation_dictionary, lookup_table, lookup_key, condition_tag, condition_value.

Used by keyword_extract derivation. Edit a dictionary's entries below — keys are output values, values are arrays of keywords that match (case-insensitive, longest-match-wins).

Sample tags with blank values are the targets — fill in the auxiliary tags so derivations have something to work with.

enabled, auto_merge_threshold (≥ this → auto-apply mapped MRN), flag_threshold (≥ this and < auto → push to manual queue), strip_leading_zeros, mrn_prefix, mrn_suffix, name_similarity_min, dob_exact_match, dob_tolerance_days, require_sex_match, unmatched_policy (flag / reject / accept), match_order (array of strategy names in priority order).

Registry seeds via Seed from archive (one-shot QueryStudies grouped by PatientID, denormalized name only; DOB/Sex backfill as ingest hits accept policy).

When a scan is running, every inbound C-STORE / DICOMweb STOW dataset is observed and counted (no pipeline change). On Complete, the scanner generates per-tag stats + auto-suggested crosswalk mappings + a list of data-quality issues. Populate crosswalks pre-fills the Phase 2E tables with discovered values (blank mappings; operator fills the destinations).

Schema: id, name, priority (lower = earlier), enabled, description, conditions[] (tag, operator: is_blank/is_not_blank/equals/not_equals/contains/not_contains/starts_with/ends_with/matches_wildcard, value, case_insensitive), actions[] (tag, action_type: set/copy/prepend/append/regex/clear/lookup, value, value2, lookup_table). Phase 2C = CRUD only. Engine apply is Phase 2D.

JSON dict of { "KeywordOrHex": "value", ... }. Keywords use the same fallback chain as the engine (preset -> DicomTag.Parse hex -> fo-dicom keyword). Engine applies every enabled rule in priority order against a per-rule clone of this dataset; the table below lists rules that matched + their changes.

Schema: name (filename slug; no spaces), display_name, description, dicom_tag (the DICOM keyword whose changes auto-record into this table — e.g. AccessionNumber), hl7_field (informational), auto_capture (bool), case_sensitive (bool), default_value, entries (object: { "old": "new", ... }; wildcards * / ? supported).

Two-column CSV: old_value,new_value per line. Header rows starting with old / source / from / # are skipped.

Schema: id, name, description, enabled, priority (lower = earlier), on_fail (Pass / Warn / Hold / Reject), check_type (required / not_empty / min_length / matches regex / value_in / not_blank_when), tag (DICOM keyword), value (predicate parameter), valid_values (array, only for value_in).

JSON dict of sample DICOM tags. Gates evaluate against this synthetic dataset; no persisted state touched.

Datasets that failed an in-flight transform (Validation / MissingData / Reconciliation / QualityGate) are persisted here. Retry re-feeds the saved bytes through the full pipeline so fixed rules immediately apply.

Inbound HL7 ORU reports are matched to archived studies by accession number, then forwarded to a destination VNA sibling per its configured policy. Records destined for an unconfigured destination are blocked until the operator picks a forwarding policy.

Status counts

Records

TCP MLLP receiver per HL7 v2.x. Default OFF; port 2575 is IANA-registered hl7 for non-TLS MLLP. Plain TCP only in this release.

Connection string configured

—

HL7 inbox catalog database. Defaults to the archive catalog engine (same connection string) unless explicitly overridden. Engine + connection string changes require service restart.

Watches {dataDir}/<relative-path> for *.hl7 / *.txt files. Picks up, parses, routes through the HL7 sink, and moves to processed/ or failed/. No subfolder recursion. Toggle requires service restart.

POST /api/hl7/ingest accepts multipart upload or raw text paste. Enable toggle and max upload size apply immediately. Retention days are read live by the manual purge endpoint.

Paste a raw HL7 v2 message; the tool parses + picks the matching per-sender mapping (or defaults) + extracts every field. Useful for verifying a new override before saving + for debugging "why is accession empty when the message clearly has one?"

Generates the configured number of synthetic studies for the chosen modality and feeds every instance through the local IInstanceReceiver — exercising every transform (Validation / MissingData / Reconciliation / TagMapping / Crosswalk auto- capture / QualityGate) + the archive write. Synthetic sourceAe = MIGRATION_TEST so audit can distinguish test from real-PACS source. Capped at 1000 studies per run.

Last run result

Stream catalog rows from a source database to a target. Both endpoints must be reachable + the schema must already exist on the target (run the host's schema-init separately).

v1 limits: Postgres ↔ Postgres fully supported. SqlServer + Oracle target stores will throw at UpsertInstance until those impls ship in separate DCRs; the job surface reports the throw message as last_error. PatientBirthDate + PatientSex are not preserved in v1 (denormalized study row carries only PatientId + Name).

New job