Skip to content

ClearPlan v3.2.0 - Configurable review and synthetic Technical Note

Latest
Latest

Choose a tag to compare

View all tags
@Kiragroh Kiragroh released this 11 Sep 03:10
· 6 commits to main since this release
v3.2.0
d21dbd5

ClearPlan v3.2.0

Software and author-review Technical Note materials are associated with the
v3.2.0 release.
The verified application source revision is
ca3e381813489542bae4551db9db23c8c3da23d8; publication-only changes add manuscript,
figures, evidence and document-generation tooling. This is not an accepted paper,
a clinical installer, or clinical commissioning. Historical v3.1.0 records are
retained separately rather than reused as evidence for this version.

Scope

  • Shared read-only plan review with safer structure mapping, explicit target
    coverage, focused interactive DVHs, shared structure visibility and default
    hiding of unmatched rows without removing the mapping editor.
  • Target-specific PAM with explicit single-/dual-layer MLC profiles, Paddick CI
    and its reciprocal, GI/HI, total MU and separately labeled estimated dose-rate
    trajectories. Unsupported or incomplete inputs remain unavailable.
  • A separate comparison reference, three CT planes with dose/structure overlays,
    optional field-start BEV/DRR panels, compact active-plan PDF and offline HTML
    quicklook. The active-plan report is not a comparison report.
  • Versioned external configuration for aliases, defaults, field nomenclature,
    constraints, target rules and enabled checks, including author/time/hash
    history and restore-as-new-revision.
  • Seven retained baseline/fault scenarios plus two generated publication
    phantoms. The latter share analytic anatomy/dose definitions across displays
    and numerical checks; dose is not calculated from the MLC apertures.
  • Optional, disabled-by-default ARIA report-document transfer, separate from
    read-only treatment-plan access. Sparse readback may use explicitly configured
    restricted UNC roots to compare the stored PDF bytes; metadata alone is not
    full verification. Local configuration and interoperability evidence remain
    private and require separate approval.

The Stock 2024 workbook remains editable example configuration, not a clinical
recommendation: its current inspected inventory is eight tables, 495 executable
rules and 70 canonical structures. Table confirmation and local review are
required. The smaller original example workbook remains available.

The illustrative RayStation adapter demonstrates the neutral snapshot boundary;
it is not a commissioned implementation of every metric. This
repository's Essentials wrapper is also distinct from a separately deployed
institutional Runner Hub.

Reproduce the software checks

Follow reproducibility. The public entry point accepts
an explicit Python executable:

.\tools\build-public-release.ps1 -Version 3.2.0 `
  -PythonExecutable python

Run under an approved execution policy. -DevelopmentPreview and
-SkipSimulatorSmoke are explicitly incomplete verification modes, not the final
release gate. No proprietary TPS runtime is required for these public tests.

Core/contract, locked synthetic DICOM, illustrative RayStation and stock tests
are separate evidence scopes. Their actual totals, failed/skipped counts,
timestamps and tested-input hashes belong in the current execution receipt, not
as fixed numbers copied into this page. The private stock-source transcription
class is absent from a clean public checkout and is explicitly skipped; parser
tests and the distributed-XLSX loader regression still execute.

Evidence and publication boundary

Use the release manifests and receipts for exact source commits, artifact
hashes, file inventories and measured environment versions. No older version's
hash or workstation record is a substitute.

  1. Reviewed public source/configuration manifest with exact-byte hashes, original
    licenses and a documented relationship to the evaluated checkout.
  2. Fresh public build and execution receipt
    (artifacts/public-release-test-evidence.json), including separate suite
    results and honest skips. Working-tree changes must be accounted for by the
    source manifest rather than concealed behind a HEAD commit label.
  3. Verified simulator runtime/package inventory and matching hashes from repeated
    packaging. Repackaging one build is not an independent clean-build experiment.
  4. A separate patient-free publication manifest: reviewed GUI figures, PDF/HTML
    examples, manuscript, references and traceability. Software-test success does
    not imply visual QA or manuscript acceptance.
  5. Independent readback of the newly published tag/commit and downloaded assets,
    matched to their final manifests and SHA-256 values. Release publication is
    not complete merely because source metadata names v3.2.0.

The paper's paper/evidence/technical_note_evidence.json records an actual run
from the frozen software revision. Its collection-time publication state is kept
unchanged; the collector does not attest to a later GitHub publication. The
release's separate publication/readback manifest supplies the tag and asset
bindings. Protected native snapshots, reports,
FHIR receipts, source workbooks and local credentials never enter the public
source or publication manifest.

Acceptance boundary

No clinical sensitivity, specificity, delivered-dose accuracy or universal
constraint applicability is established by the synthetic fixtures. Native TPS
interoperability, installed machine profiles, ARIA transfer and clinical workflow
acceptance require local controlled verification. Use the
commissioning scaffold,
analysis method and
ARIA guide as scoped implementation aids, not as
clinical approval.

Assets 13
Loading