Quick start
Define the case. Set PI, LL, TK (T4–T12), CL and PT with the sliders, or open Examples & import and load an example. Coronal inputs and measured SVA sit under Additional measurements .Review the anatomy. Rotate the model, use the Lat / AP / PA / Obl presets (keys 1–4), and switch between Pre-Op and Modeled Post-Op .Build the procedure. Pick a level and a procedure, check the planned change, then Add to itinerary . Use Edit on a card to change it, and the Review tab for the modeled output.
Undo any case change with Ctrl+Z (Ctrl+Shift+Z to redo). Save the case as a .json file from the folder button in the header, and open it later or drop it onto the workspace. Everything runs on this device.
Keyboard and mouse (3D viewport focused)
Drag / one-finger drag Rotate Scroll / pinch, + and − Zoom Right-drag / two-finger drag Pan Arrow keys Orbit 1 · 2 · 3 · 4 Lateral · AP · PA · oblique view R or 0 Reset the camera to the selected view Double-click a level Isolate it (Esc to exit) Ctrl+Z · Ctrl+Shift+Z Undo · redo a case edit (outside text fields)
Model card · Limitations · Evidence
What the simulator shows
The reference scene is generated as a parented kinematic chain from the feet through the skull. The skull, intact C1–L5 levels, sacrum, and bilateral hip bones use embedded CT-segmentation-derived external surfaces from one de-identified TotalSegmentator v2 source subject. Complete femur-to-metatarsal standing context uses a separately sourced BodyParts3D bilateral reference atlas. Each source is uniformly fitted without changing surface proportions; this is deliberately labeled composite reference anatomy, not one coherent patient. A pelvic incidence other than the atlas value is shown by rotating the sacrum and spine about the L5–S1 joint while the hip axis stays fixed; the hip bones are then reshaped schematically so the sacroiliac joints and hip sockets stay in contact, an educational approximation rather than patient anatomy.
Alignment engine
Modeled LL and CL add explicit level-specific scenario corrections. C2–C7 lordosis excludes C1–C2 and regional curves use declared, normalized display profiles rather than equal rotation at every joint. Thoracic osteotomy correction changes TK directly. The reciprocal thoracic response is off by default; when enabled, TK increases by 0.2° per degree of lumbar gain. The pelvic response is always applied when lumbar lordosis is gained — PT decreases by 0.5° per degree gained at L4–S1 and by 0.25° per degree at L1–L4 and the T12–L1 junction (level-dependent coefficients from Lafage R et al., Global Spine J 2022, PMID 33567927), floored at the lower of 10° and the PI-relative ideal PT — because compensatory pelvic retroversion is the posture a hypolordotic lumbar spine forces, and holding it after lordosis is restored would rigidly lever the modeled trunk backward into a stance no standing person could hold. That is kinematic bookkeeping for a standing chain, not an outcome prediction, and the displayed PT is tagged as a balance response. The app does not calculate SVA or a synthetic balance score.
The coronal/axial layer accepts an authored primary Cobb surrogate, secondary Cobb surrogate, apical translation, axial rotation, and operation-specific signed changes. It does not reconstruct Cobb end vertebrae or infer flexibility from imaging. Optional secondary-curve response is bounded by the fraction of primary-curve improvement and a user-declared 0–100% flexibility value: 0% leaves a structural curve unchanged, while 100% permits the full bounded proportional response. The model is an editable kinematic scenario—not real physics, a force solution, or a patient-specific prediction.
Anatomy, ligaments, and resections
The embedded CT surfaces provide whole-bone external shape but do not contain validated pedicle corridors or semantic lamina/facet partitions. Laminectomy and SPO therefore retain the intact CT whole-bone surface and add source-bounded planned posterior resection-envelope and cut-margin overlays. PSO likewise retains the CT surface and adds a translucent, source-bounded planned three-column wedge envelope. These overlays are not CT mesh subtraction, do not claim a completed patient-specific resection, and do not infer cut safety. TOPS automatically adds its same-level decompression item and, in the modeled post-op view, replaces the two intact display surfaces with derived open post-decompression shells that remove source-bounded posterior triangles through the lamina/spinous/facet display regions while preserving the original CT meshes for pre-op. Those shells demonstrate the decompression footprint but are not Boolean solids, patient-specific cut plans, or clearance validation. The external TOPS display follows the public two-crossbar, four-screw, compact-boot topology; proprietary internal geometry and patient fit remain withheld. Lower-limb hip, knee, and ankle pivots use documented surface-geometry heuristics only—not validated functional joint centers. Ligaments are thin, CT-envelope-guided attachment/pathway proxies that meet at shared motion pivots; they are not segmented tissue or force-bearing elements.
Ligament pathways, SI surfaces, and alignment measurement guides are overlays, not bone anatomy; they are hidden by default and can be enabled from Layers. Implant shafts, rods, cage envelopes, and arthroplasty envelopes use explicit catalog millimetre dimensions selected by the user and converted through the shared scene scale. Catalog availability is not a sizing recommendation. Screw corridors, containment, interference, fit, head/thread details, cage teeth/porosity, and graft packing are not validated. Components without a dimensioned catalog fail closed and are not replaced by generic oversized geometry.
Third-party names. TOPS, XLIF, OLIF25, Mobi-C, HEDRON, neon³, uCentum, uBase, MASTERGRAFT, ViviGen, Infuse, and other referenced product or procedure marks belong to their respective owners. References are used only to identify source material or a catalog configuration and do not imply sponsorship, affiliation, endorsement, or freedom to operate.
Anatomy Lab reference-state editor
Anatomy Lab stores manually authored three-plane shape, standing-alignment context, and level-specific reference-appearance factors separately from the alignment calculator. Loaded disc-height loss is a geometric percentage, not an MRI or Pfirrmann grade. Shape transforms alter only simple display proportions; they do not recreate congenital topology, establish enumeration, or diagnose a variant. When an intact CT reference surface is available, degeneration and component-appearance factors never replace it with a generic whole-bone mesh; unsupported component shape edits are withheld while supported disc and bounded overlays remain visible. Exploded inspection is non-anatomic and disables measurement, containment, and collision interpretation. Cervical arthroplasty represents disc replacement after discectomy and endplate preparation; TOPS represents posterior-element/facet replacement after decompression, not disc replacement.
Evidence-linked construct comparisons
The comparison panel matches transparent authored patterns—such as positive sagittal alignment, regional coronal/axial factors, cervical or thoracic deformity, focal junctional angulation, and lumbar level-specific collapse or translation—to unranked construct scenarios. Each scenario exposes why it appeared, editable representative branches or sequences, linked study context, and the patient-specific inputs still missing. Prefill changes only the procedure editor; it never adds a procedure or claims that a construct is indicated, preferred, feasible, or likely to achieve the seeded correction.
Postoperative construct review
The Modeled Post-Op view first aggregates every same-level procedure and the saved implant-plan components; a missing implant-plan snapshot fails closed and is not described as actual fixation. ALIF, LLIF, XLIF, and OLIF default to a cage plus bilateral posterior screws and rods, while an explicit editor control can author an interbody-only variant. A cage plus posterior screw–rod components is treated as a posterior-supported planned fusion—not as an untreated mobile segment. TOPS after the required laminectomy receives a separate motion-preserving-device label rather than a fusion label. Blue markers identify the first segment outside a planned fusion span; teal markers identify a posterior disruption paired with authored fixation or a motion device; amber markers identify an authored posterior disruption without either. These are topology and factor-presence labels only. They do not measure motion, prove fusion, diagnose instability, or calculate a probability.
Mechanics validation gate
The app may identify mobile levels adjacent to a modeled fusion, but it intentionally withholds implant stress, fatigue life, screw loosening, cage subsidence, ligament strain, adjacent-segment degeneration, and failure probability. Those outputs require patient/reference-specific geometry, nonlinear tissues and contacts, bone density, implant dimensions and materials, fusion state, boundary loads, mesh convergence, uncertainty analysis, and independent experimental validation. A render mesh is never treated as the finite-element model. Declared input/output ranges are engineering guardrails only—not normal ranges, treatment targets, or clinical thresholds; scenarios outside them suppress modeled alignment rather than extrapolate.
Model card · v0.23.1 · alpha
Intended use: educational and research scenario exploration. Angles and optional measured SVA are user-entered; every correction records whether it is user-planned or a procedure-default assumption; the reciprocal thoracic response is visibly opt-in and assumption-derived, and the always-applied pelvic balance response is tagged as such on the pelvic-tilt row. Clinical performance has not been established, and the engine must not be used for diagnosis, treatment selection, implant sizing, or patient-specific outcome prediction.
Startup model integrity: checking…
Patient imaging boundary
The local intake accepts only institutionally de-identified research sources. For browser-decodable standing raster images, an experimental local auto-landmark beta can estimate a spine corridor, selected endplate lines, and radiographic measurements; every result remains review-required and unvalidated. User-accepted measurements can be transferred to the reference-atlas alignment and immediately visualized, but the uploaded images never become exact 3D anatomy. CT-to-mesh reconstruction remains disabled and requires segmentation, level numbering, per-vertebra standing registration, and human QA in a validated deployment. HIPAA compliance requires administrative, physical, and technical safeguards beyond this file (HHS Security Rule guidance ). De-identification must follow an approved method (HHS de-identification guidance ); DICOM confidentiality profiles alone do not guarantee removal of private tags, burned-in pixels, or recognizable facial anatomy (DICOM PS3.15 Annex E ).
SRS–Schwab reference modifiers
PI–LL, measured SVA, and PT are shown separately as 0 / + / ++ reference bands only in the pre-op view. SVA remains unavailable unless an SVA input is present, and all bands are unavailable for modeled post-op values. The bands are not combined into a diagnosis or target; the full classification also requires a coronal curve type and clinical context.
Modifier Grade 0 Grade + Grade ++
PI–LL < 10° 10–20° > 20°
Measured SVA < 4 cm 4–9.5 cm > 9.5 cm
PT < 20° 20–30° > 30°
Evidence & governance
Three-plane context: adult deformity literature distinguishes comparatively rigid structural curves from typically smaller, more flexible compensatory curves (adult spine deformity review ) and emphasizes assessing curve flexibility rather than assuming spontaneous correction. Reciprocal whole-body changes after realignment are documented but heterogeneous (whole-body reciprocal-change review ). Lateral interbody surgery can change coronal and sagittal alignment in cohorts (systematic review and meta-analysis ), but this app intentionally applies no universal procedure-to-correction ratio.
Construct-comparison context includes an ALIF registry comparing stand-alone and posterior-fixation constructs , a three-plane lateral interbody cohort , SPO/PSO correction-range review , an adult cervical deformity approach cohort , MISDEF2 strategy/outcomes data , and the TOPS randomized FDA IDE comparison . Study populations and limitations are displayed inside every surfaced option; none is converted into a treatment rank or patient-specific correction estimate.
Implant catalogs: HEDRON C/A/L/P/T interbody size matrices , FDA Mobi-C labeling , neon³ cervical fixation catalog , and uCentum/uBase thoracolumbar and pelvic fixation catalog . Graft selectors represent manufacturer package quantities from FDA AccessGUDID, LifeNet Health, and Medtronic; package volume is never inferred to equal volume placed at the selected level.
Anatomy QA references include adult occipital-condyle CT morphometry , C0–C2 MDCT joint intervals , level-specific cervical disc heights , thoracic disc anatomy , and level-specific lumbar CT disc heights . These ranges audit the reference; they do not rescale individual bones or imply patient-specific validation.
Geometry provenance: TotalSegmentator v2 small dataset (subject s1369; CC BY 4.0) and its peer-reviewed dataset paper . Lower limbs are adapted from BodyParts3D v4: “BodyParts3D, © The Database Center for Life Science licensed under CC Attribution 4.0 International.” The current archive license page says CC BY 4.0, while the supplied OBJ headers retain a legacy CC BY-SA 2.1 Japan notice; both are preserved in the build manifest and legal confirmation is required before App Store or Play Store distribution. Secondary audits include the public-domain NLM Visible Human Project . Modeling references include C3–L5 full-spine CT morphometry , ligament-informed lumbar calibration , the SRS–Schwab classification overview , an Ames cervical-classification review , and standing cervical/global alignment by Roussouly subtype . Render meshes are not finite-element meshes.
Postoperative factor context is drawn from peer-reviewed evidence on proximal junctional kyphosis factors , a separate PJK/PJF meta-analysis , the lumbar adjacent-segment systematic review , and a human cadaver laminectomy study . These sources justify showing factor presence; they do not validate this app as a patient-specific prediction model.