Orcaton
Sign in
Spine SABRintermediateAI-generated CT

Synthetic Spine SABR Planning

30 Gy in 3 fractions·contributed by Orca Image·2026-08-27

This CT was generated by a model. No patient was scanned.

The anatomy is synthetic and the structures come from the generator rather than from a clinician. Treat the Hounsfield units as plausible rather than measured: there is no scanner behind them and so no CT-to-density calibration of yours applies, and because the image has no real counterpart the error in them cannot be quantified at all. It is a planning exercise, and every plan here is scored against the same geometry — but nothing learned from this case is a statement about a real patient.

  1. 1

    Get the case

    A de-identified CT and structure set, as one archive. Plan it in your own system.

    No archive attached yet.
  2. 2

    Plan it

    The criteria are published before anyone plans and do not change mid-challenge. Below they are also written as entries for your own optimizer.

  3. 3

    Submit the dose

    Upload the RTDOSE. It is scored automatically against the criteria, and you are told when it lands.

Guideline

The rules that decide whether a plan counts, before the numbers that score it.

Maximise target coverage while strictly respecting the neural constraints. The neural limits disqualify; everything else is scored.

Based on KROG 13-16, Spine SBRT for metastatic epidural spinal cord compression (multicentre prospective phase II). Dose ranges, neural and abdominal constraints, and the unacceptable-plan criteria are taken from it.

Where the points are

Target coverage35
Neural and other OAR sparing35
Conformity, homogeneity, gradient15
Deliverability and complexity15

100 points published. The scoring criteria below are what this platform computes automatically.

Allowed

Optimisation structures
Rings, PRVs, target-minus-organ — build whatever your system needs. Scoring reads the published contours, never yours
Modality
Photon only
Treatment unit
Linac-based — C-arm, Halcyon, TomoTherapy
Energy
Any clinically commissioned photon energy
Technique
VMAT, IMRT, SRS/SBRT arc, or helical
Dose calculation
Heterogeneity-corrected algorithm
Dose grid
≤ 1.5 mm recommended, 2.0 mm maximum
Normalisation
Free, provided the prescription and DVH reporting stay consistent

Not allowed

Modality
Proton, particle, electron, brachytherapy
Contours
Editing the provided GTV, PTV or OAR contours
Deliverability
Plans that could not be delivered clinically
Density override
Unless explicitly permitted
Bolus
Unless explicitly permitted

What disqualifies a plan

A fail removes the plan from the ranking whatever it scored. Some of these are checked from the dose automatically; an edited contour or an undeliverable plan is not, and is enforced by review.

  • failNon-photon or non-linac plan
  • failPrescription other than 30 Gy in 3 fractions
  • failParticipant-edited GTV, PTV or OAR contours
  • failMandatory neural constraint exceeded
  • failPTV V30Gy < 80%
  • failGTV V30Gy < 95%
  • failNon-deliverable plan
  • majorDose grid larger than 2.0 mm
  • majorHot spot outside the target
  • majorAny serial OAR mandatory D0.03cc exceeded

What to submit

  • requiredRTDOSE
  • recommendedRTPLAN
  • recommendedDVH report
  • recommendedAxial / sagittal / coronal isodose screenshots
  • optionalPatient-specific QA result

Worth knowing before you start

  • Deliverability — monitor units, modulation and treatment time — is 15 of the published 100 points and is not scored automatically. RTPLAN uploads make the MU part measurable; the rest is assigned by review after the deadline, as the original challenge did.
  • The source document marks GTV V30Gy < 95% as 'major violation or fail'. It is enforced here as a fail, which is how the original challenge ran it. A plan at 94.9% is disqualified — worth knowing before spending the run on an organ at risk.
  • The guideline recommends ranking conformity and gradient by percentile among valid submissions rather than fixed thresholds, because both are sensitive to target size and level. Fixed goals are used here so a plan can be scored on arrival; the percentile view belongs on the leaderboard, not in the score.
  • Where the guideline writes '≤ 22 → 20' it is recording a revision. The later figure is used.

Scoring criteria

85 points

How a number becomes points. A metric at or past Ideal earns its full points. At the Minimum it earns none, and between the two it is linear — halfway between minimum and ideal is half the points. Worse than the minimum is zero for that line, never negative, and the rest of the plan still scores. Breaching a hard constraint is different: the plan keeps its number and leaves the ranking. How scoring works in full →

Target

MetricMinimumIdealPoints
GTV V30Gy 95 9810
PTV V30Gy 90 958
PTV D95% 27 307
PTV D90% 27 305
PTV Dmax 115 1055

Organs at risk

MetricMinimumIdealPoints
Cauda equina D0.03ccon CaudaEquina 24 2012
Cauda equina D0.35ccon CaudaEquina 22 186
Cauda equina D10%on CaudaEquina 22 185

Organs at risk — sparing factor

Each group starts at its weight and gains a point per Gy spared below the ideal, so sparing further keeps paying off.

  • Bowel, stomach and duodenum6 pts

    d0.03cc ≤ 24, d5cc ≤ 14.5, d0.03cc ≤ 22, d20cc ≤ 18

  • Kidney3 pts

    d200cc ≤ 12, d200cc ≤ 12

  • Skin and other3 pts

    d0.03cc ≤ 22, d10cc ≤ 20, max_dose ≤ 24

Plan quality

MetricMinimumIdealPoints
CIon PTV 1.2 1.055
HIon PTV 0.4 0.255
GIon PTV 6 35

Hard constraints

Breaching any of these removes the plan from the ranking, whatever it scored.

  • Cauda equina D0.03cc > 24 Gy (mandatory neural constraint)
  • Cauda equina D0.35cc > 22 Gy (mandatory neural constraint)
  • PTV V30Gy < 80%
  • GTV V30Gy < 95%

Spine SABR Plan Challenge Guideline (v2), adapted from KROG 13-16 by AOCMP 2026 Plan Challenge education committee.