DICOM Migration Reconciliation: How to Prove Every Study Arrived

What migration reconciliation means

DICOM migration reconciliation is the controlled comparison of the agreed source inventory with evidence from the destination. Its purpose is to account for every expected study and to expose missing, partial, duplicate or unexpected outcomes. It is different from watching a transfer queue: operational completion describes process activity, while reconciliation tests whether the agreed data outcome was achieved.

Create an authoritative study ledger

Use one study-level record as the control ledger throughout the project. The Study Instance UID is normally the strongest DICOM identifier for matching, supported by Patient ID, accession number, study date and modality for investigation. Preserve the original discovered values even if destination demographics are corrected later.

• Expected Study Instance UID and source system.
• Patient ID, accession number, study date and modality.
• Transfer state, attempt history and last response.
• Destination verification state and exception disposition.

Use unambiguous outcome states

A single completed flag is too coarse for migration acceptance. Define states that reflect both execution and verification so an operator can distinguish a successful request from a verified destination study.

• Discovered: included in the authoritative scope.
• Requested: a transfer operation was initiated.
• Received: destination evidence confirms the study.
• Verified: identifiers and agreed completeness checks passed.
• Exception: a failure or discrepancy requires action.
• Approved exclusion: explicitly removed from acceptance scope with reason.

Compare identifiers, not totals alone

Source and destination totals are useful as an early warning, but equal totals do not prove equality. One missing study and one unexpected study can leave the total unchanged. Perform set-based comparison using Study Instance UID, then use secondary identifiers and instance counts to investigate mismatches.

• Expected but not found at destination.
• Found at destination but not present in the expected set.
• Identifier conflict or demographic mismatch.
• Study present with a materially different instance count.
• Duplicate or merged records requiring documented treatment.

Detect partial and clinically unusable studies

Study-level presence may still conceal missing series or instances. Where the project risk requires it, compare series and SOP Instance UID counts or hashes available from the systems. Validate representative studies through the clinical viewer, because database presence alone does not prove that users can retrieve and display the images.

Investigate and close exceptions

Classify each discrepancy before retrying it. Common causes include inaccessible source objects, incorrect AE routing, destination rejection, unsupported transfer syntax, timeout, duplicate handling, demographic rules or an inventory created with inconsistent query criteria. Preserve the technical response and the decision that closed the exception.

• Retry after a transient infrastructure fault.
• Remediate configuration or data, then reprocess.
• Accept a documented exclusion approved against project criteria.
• Keep unresolved discrepancies open; do not fold them into completed totals.

Build the migration evidence bundle

The evidence bundle should allow an independent reviewer to understand what was expected, what happened and what remains. Use stable exports and checksums where possible, and record the software version and query criteria used to create each report.

• Approved scope and source inventory.
• Destination verification extract and reconciliation report.
• Exception register with action, owner and final disposition.
• Operational summary, transfer logs and pilot evidence.
• Clinical sampling results and formal acceptance record.

Report metrics that support decisions

Report expected, verified, approved-exception and unresolved counts separately. Add completion percentage only when its denominator and rules are explicit. Operational throughput, failure rate and retry rate help manage the project, while verified-study coverage and unresolved exceptions support acceptance decisions.