The case for analysing the patient’s own bone.

Most implant analysis, when it happens at all, is run against a generic or scaled bone rather than the patient in front of the surgeon. Bone is not a uniform material: a dense cortical shell surrounds a porous trabecular core, and the balance between the two varies by anatomical site and by patient — most visibly at the metaphysis, where the cortex thins and the trabecular structure carries a larger share of the load.

That variation is exactly where the orthopaedic literature places several of the mechanisms associated with implant revision — aseptic loosening, stress shielding and the adaptive bone remodelling it drives (Wolff’s law), periprosthetic fracture, and micromotion at the bone–implant interface sufficient to prevent osseointegration. These are described here as mechanisms discussed in the literature, not as predictions about any patient — an averaged, template-based model has little to say about any of them, because they are governed by exactly the local variation it averages away.

Engineers working in orthopaedic biomechanics rarely have a direct, repeatable path from a patient’s clinical imaging to an analysis-ready model. Clinicians rarely have access to patient-specific engineering analysis run against a case they actually hold — the CT sits in one system, the analysis capability in another, and the two are brought together by hand, if at all.

That gap — not a shortage of modelling technique on either side — is what a shared, repeatable surface between imaging and analysis closes. It is also why this platform is built to be legible to both a biomechanical engineer running the analysis and a surgical or radiological collaborator reading the imaging behind it, rather than to only one of them.

The model is only as good as the scan it comes from. A usable CT needs slice resolution and coverage sufficient to resolve the region of interest, and ideally a calibration reference in the field of view so density values map to real bone mineral density rather than an assumed scanner curve.

The honest caveat, stated rather than avoided: where an existing implant is already present, metal artefact degrades image quality in the region nearest it — precisely the revision case where a patient-specific model is most valuable. Where that affects a given case, it is reported as a known limitation of that case, not implied away.

Every case runs through 76 automated checks before a result is reported, and every run records what produced it — so a result can be regenerated, not just read once and trusted. That reproducibility, more than any single number, is the engineering claim this platform makes.

The framing follows ASME V&V 40 — the standard for assessing the credibility of computational models used in medical device contexts — because it is the vocabulary this field already uses to ask "how much should this model be trusted, and for what." And the absence of regulatory status is a positioning, not an apology: as engineering software for design and research, it carries no clinical clearance and needs none — computational analysis of this kind is routine in device development, used to inform a design or research decision rather than to diagnose or treat.

The design that ends up correct is rarely the first one tried. A parametric sweep across size, thickness or position runs every variant as a complete, fully solved analysis — not an estimate interpolated between two that were actually run — and returns the results side by side for comparison.

Measured the way an engineering team measures it: in design iterations, not in money or clinical outcome. A design question answered while the geometry is still a parameter is cheaper to answer than the same question asked of a physical prototype.

Typically: CT data already collected for clinical purposes, and a short conversation about the case and how the analysis result will be used. For a first case, that is roughly an hour of clinical time — reviewed up front, not discovered partway through.