SPIN-THA v7.6 — Spinopelvic planning support

Boundary

Surgeon-supervised planning support. Not a diagnostic device.

SPIN-THA is positioned as planning support under a surgeon's supervision. It does not recommend a cup orientation, does not diagnose instability, and does not replace surgeon judgment. The embedded PelvisAtlas backbone (TCIA CT-LYMPH-NODES, a healthy lymph-node imaging cohort, N=12) is used as shape context only — not as a population reference for clinical decision making. See docs/V7_5_FDA_CDS_ASSESSMENT.md and docs/V7_5_PHI_POLICY.md for the full positioning.

01 — Engine outputs

Eleven outputs.
No black boxes.

The deterministic engine produces a fixed set of typed outputs from a fixed set of typed inputs. Every coefficient regime, every band, every fallback is named and visible.

Rule-state signal

Non-numeric level — high / watch / low — with the list of named, plain-text drivers that fired. The legacy uncalibrated "risk score" is gone.

Age-adjusted PI–LL band

Lafage-style age-adjusted band rather than a flat |PI−LL| > 10° rule. Driver: pi_ll_age_adjusted_mismatch.

Approach-conditional combined-version band

combinedVersionBandFor(case) returns posterior, direct-anterior, lateral, and default bands. The approach changes the band, not the patient.

Rotation-matrix cup orientation

Geometric rotation-matrix cup orientation is emitted alongside the standard small-angle linear approximation, so the reviewer can see where they diverge.

Tilt-coefficient regime

Coefficients are case-conditional on the AV definition (radiographic / operative / anatomic). A coefficient_regime_warning driver fires when |ΔPT| exceeds 25°.

Plausibility gate

plausibilityIssues() against PHYSIOLOGIC_RANGES. Negative ΔSS and similar physical impossibilities raise an implausible_measurement driver (HIGH).

±2°/±5 mm sensitivity

Per-input sensitivity analysis around the entered values. Surfaces which inputs the rule-state actually depends on for this case.

Monte Carlo uncertainty

Uncertainty propagation under per-input observer confidence. Missing confidence entries are excluded, not imputed to 0.5.

T1-pelvic-angle proxy

Computed T1PA proxy emitted alongside the spinopelvic block. Useful when a measured T1PA isn't on the radiograph.

FHIR R4 Bundle

A fhir_bundle of Observations covering every present measurement and every derived metric. One-click Download FHIR Bundle action.

DICOM SR (TID 1500)

Structured report following TID 1500 with TID 5200-style measurement groups, linked to the source imaging study by SOP Instance UID.

Ed25519-signed manifest

The reproducibility manifest is signed with Ed25519. Hash is over inputs only: case_id, atlas_id, engine_version, rule_version, measurements, history, imaging, cup_plan.

02 — Validation

Tested against a 24-case
synthetic bank.

The engine is validated against a synthetic case bank covering three mobility phenotypes, with the audit history bundled in the repository. 33 of 38 findings from the independent v7.4 review are addressed in the shipped v7.5/v7.6 build.

Phenotype A — STIFF

8 cases · low spinopelvic mobility

Phenotype B — ADAPT

8 cases · adaptive intermediate mobility

Phenotype C — HYPER

8 cases · hypermobile spinopelvic relationship

Audit findings

33 of 38 shipped · history in docs/

Smoke tests

run_engine_tests_v7_5.js

Package verifier

verify_package_v7_5.py

03 — Example manifest

Inputs-only hash.
Replayable forever.

The reproducibility hash is over inputs alone — feed the same inputs back into the same engine version and you get the same outputs byte-for-byte. Engine version is pinned globally as window.SPINTHAEngineV75.

// reproducibility manifest (excerpt) — SPIN-THA v7.6 / engine v7.5
{
  "case_id":        "SPINTHA-ADAPT-002",
  "atlas_id":       "PelvisAtlas/TCIA-CT-LYMPH-NODES@N12",
  "engine_version": "v7.5.0",
  "rule_version":   "v7.6.0",
  "measurements": {
    "PI": 54.2, "PT": 12.6, "SS": 41.6,
    "LL": 48.0,
    "cup_inclination": 44,
    "cup_anteversion": 22,
    "femoral_version": 15,
    "av_definition": "radiographic"
  },
  "session_token":         "fnv:a13c…ef",
  "sha256_canonical_input":"sha256:1ab1ee…3625",
  "signature_alg":         "Ed25519"
}
    

04 — Operating it

Built for the desk
between two patients.

Keyboard-first navigation, undo on the last measurement, local persistence so a power blink doesn't cost a case. Print is gated behind an explicit confirm dialog.

←/→

Prev / next stage

Move through the workflow without taking a hand off the keyboard.

1–9

Jump to stage

Direct numeric stage selection from anywhere.

Undo last edit

Reverse the most recent measurement entry. Same as the explicit Undo button.

⌨

Local persistence

Case state, edits, cup adjustments, validation entries, observer ID, and active step persist via localStorage and IndexedDB.

05 — Companion materials

A separate anatomical viewer
plus a SoftwareX paper.

The single-file 3D pelvis viewer ships alongside the workbench. The SoftwareX paper, BibTeX, figures, and results table live in manuscript/ in the repository.

VIEWER

SPINTHA Anatomical Viewer

Single-file 3D pelvis viewer. No build step. Same offline guarantee as the workbench.

PAPER

SoftwareX manuscript

Open, deterministic, audit-driven web toolkit for spinopelvic planning support in THA. Five figures, synthetic case bank, full results table.

06 — Authors

A small clinical team.

Bone and Joint Reconstruction Research Center, Iran University of Medical Sciences, Shafa-Yahyaeian Orthopedic Hospital, Tehran, Iran.

CORRESPONDING

Behnam Sour, MD

Bone and Joint Reconstruction Research Center, Iran University of Medical Sciences. ORCID 0009-0002-3176-1956.

CO-AUTHOR

Kaveh Gharanizadeh, MD, Prof

Bone and Joint Reconstruction Research Center, Iran University of Medical Sciences. ORCID 0000-0001-6869-0333.

← All capsules Sagittal Atlas → Verify a release