Research software for adult spinal deformity
Zyra turns the standing radiographs you already take into a patient-specific model of the spine and its muscles, then estimates the compression and shear at the first mobile level above the construct.
The problem
Proximal junctional failure is one of the most common reasons adults return to the operating room after deformity surgery. The frameworks used to plan these operations, from SRS-Schwab to GAP and age-adjusted targets, describe the shape of the corrected spine.
Two patients with the same alignment can still load the top of their construct very differently, because their muscles, body mass and posture differ. That load is what the junction actually has to carry, and today nobody sees it.
How it works
Everything runs from imaging a deformity patient already has. No CT, no MRI, no motion lab.
Drop in the standing AP and lateral radiographs as DICOM, JPEG or PNG; pixel size comes from the DICOM header or a quick calibration. A detector proposes vertebral centroids from T1 to L5 on both views, and you move anything that is off.
Landmarks from both views are reconstructed in 3D. A full thoracolumbar musculoskeletal model, with hundreds of muscle fascicles, is scaled to the patient and fitted to their spine, vertebra by vertebra.
The model stands and bends through everyday tasks while the muscles do the work of holding each posture. Zyra reports compression and shear at the level above the fusion, before and after surgery, side by side.
Where it fits
First indication · in development
Run the first standing film after surgery through the model. Junctional loads are estimated for that patient, so follow-up attention can go where the mechanics say it is needed.
Next
Choose a construct and an alignment goal, let the solver fit the corrected spine, and send each candidate through the same model. Compare the loads at the junction before you choose.
The workstation
Zyra is a full planning workstation in the browser. The colours mean one thing each: cyan is measured from the film, orange is planned, and everything else stays out of the way of the radiograph.
DICOM, JPEG or PNG, with pixel size from the header or a draw-a-known-length tool.
T1 to L5 centroids proposed on both views, with every point reviewable.
PI, PT, SS, lumbar lordosis, global tilt, SVA and the GAP score, defined and documented.
Every plan is assessed against eight established alignment systems, side by side.
Pick the fusion levels and a goal; the solver fits a corrected spine to match it.
Send the pre-op and planned spine to the biomechanics engine and read the loads back.
Measurements, the plan and its alignment scores on one page for the case conference.
Single sign-on, organisation-scoped access, and an audit trail of every case read and write.
Science
Each link in the chain from radiograph to load has been tested on its own. The outcome study, which asks whether these loads separate patients who later fail from those who don't, is under peer review.
Computer Methods in Biomechanics and Biomedical Engineering. How accurate the landmarks must be for the loads to hold.
European Spine Journal. Patient-specific muscles without a CT or MRI of the trunk.
JOR Spine 8(4):e70142. The case for junctional loads, written with spine surgeons and biomechanists from several centres.
Journal of Biomechanics 188:112741. Which muscle properties change the predicted load, and by how much.
About
Zyra grew out of PhD research in biomedical engineering at the University of British Columbia on subject-specific musculoskeletal models of the spine in adult spinal deformity. The question behind it is simple: can a model built from routine imaging see the mechanical load that alignment rules miss?
Nima Ashjaee, founder. Biomedical engineer working where medical imaging, musculoskeletal modelling and machine learning meet. ashjaee.com
Research pilots
We are looking for spine surgery groups to take part in retrospective validation and research pilots. Tell us about your service and the data you hold, and we'll set up a call.