# Imaging-to-surgical-plan pairing: synthetic field guide

Version 1.0. Published October 23, 2026 by Operavue.

This original educational example contains four invented rows. It includes no scans, segmentations, actual plans, real DICOM UIDs, patient identifiers, or clinical results. Its references point to fictional artifacts that are not provided. It is not a DICOM schema, clinical standard, or automated acceptance test.

Read the worked example: https://operavue.com/insights/imaging-to-surgical-plan-datasets

## Task definition

For this exercise, the requested target is a documented plan based only on information available before surgery. The target does not require spatial geometry. Each row represents one proposed pairing, not an accepted case. Every review status is invented for the demonstration.

## Fields

| Field | Meaning |
|---|---|
| `case_ref` | Opaque project case reference. This alone does not prove linkage or anonymity. |
| `image_series_ref` | Delivered source-series reference. A real release must resolve it to the exact approved files and version. |
| `target_kind` | `documented_preoperative_plan`, `segmentation_only`, or `retrospective_reconstruction` in this example. Define a project's actual targets before collecting data. |
| `plan_ref` | Reference to the plan artifact; blank in P03 because no plan is supplied. |
| `plan_version` | Exact plan version; blank when no plan is supplied. |
| `plan_source_series_ref` | Series the source-side record says informed the plan; blank when no plan is supplied. |
| `linkage_review_status` | `reviewed`, `discrepancy_reported`, or `not_applicable` when no plan is supplied. Use `unknown` for unestablished mappings in a real review. |
| `plan_review_status` | `reviewed`, `pending`, or `no_plan_supplied`. “Reviewed” is not a clinical correctness or acceptance claim. |
| `target_evidence_basis` | `preoperative_only`, `includes_postoperative_information`, or `not_applicable` when no plan exists. Use `unknown` when the evidence basis is not established; a blank date is not proof of a preoperative-only basis. |
| `geometry_mapping_status` | All rows use `not_required_for_target`. Spatial tasks need separately reviewed mappings; unresolved is not equivalent to unnecessary. |
| `split_group_ref` | Invented grouping key. In a real experiment, define grouping using permitted, source-verified relationships among patients, cases, and versions. |
| `example_review_decision` | Educational interpretation of the declared circumstances, not a machine-produced verdict or clinical recommendation. |

## Reading the four rows

- P01: S01 is both delivered and referenced by PL01 version 1. The invented source-side review records the pairing and a preoperative-only evidence basis. Continue substantive target-quality, permissions, privacy, and release review.
- P02: S02-new is delivered, but the plan references S02-old. Resolve the discrepancy. The row does not establish that the plan is clinically wrong or that the two series are interchangeable.
- P03: segmentation is supplied without a planning artifact. Three plan fields are intentionally blank; do not fill them with zero, a guessed reference, or a negative planning label.
- P04: the reconstructed plan is explicitly stated to use postoperative information. It does not fit this exercise's target. Retrospective authorship alone is not disqualifying; a separately defined retrospective task may be appropriate.

## Before adapting this example

Define the actual target and intended model-input cutoff. Record target authorship and evidence access separately. Preserve plan revisions, reviewer roles, unresolved disagreements, and the evidence supporting the source mapping. Add a release manifest with exact artifact versions and integrity checks in the approved environment. This short educational table is not a complete delivery manifest.

The CSV cannot verify image identity, alignment, clinical validity, independence of evaluation groups, privacy compliance, or licensing rights. Check the source artifacts and supporting records through the appropriate review process. Real patient-level manifests and identifiers do not belong in public repositories, business inquiry forms, or ordinary email.

Prepared with AI assistance and reviewed for consistency with the published guide. No clinical review or endorsement is claimed. Primary references and their scope are linked in the article.