From Diagnostic Ambiguity to Pattern Clarity: Deterministic Radiographic Pattern Analytics in DISH vs Axial SpA Differentiation

RheumaView™ is a validator-governed, deterministic radiographic analytics platform designed to convert routine imaging into standardized, auditable, fixed-order outputs: structured descriptor capture, burden scoring, phenotype anchoring, and cross-modality correlation—rendered reproducibly across reruns and timepoints. This record demonstrates that architecture in a deliberately challenging axial case where real-world “label noise” and overlapping degenerative phenotypes commonly defeat non-governed workflows.

Why this case is ideal for demonstration (and why it’s hard)

• Overlapping phenotype signatures in a single patient: DISH-dominant bridging ossification + high-burden degenerative disc/facet disease + chronic SI arthropathy + hip OA.

• The classic mislabeling trap: “AS?” enters the clinical trajectory, while imaging features demand strict pattern discrimination (DISH vs axSpA vs degenerative).

• Real-world confounders (hardware and vascular calcifications) that routinely degrade scoring, segmentation, and explainability in less governed pipelines.

What RheumaView™ delivers (what typical workflows struggle to provide)

• Blinded XR-first phenotype anchoring: the initial RheumaView analysis was generated from radiographs with only minimal demographics—before post-hoc access to expanded clinical context.

• Pattern-level discrimination, not just a list of findings: explicit feature clusters supporting a DISH-dominant axial ossification phenotype with superimposed degenerative disease, while an axial SpA-dominant signature is not demonstrated on the available study.

• Structured burden scoring: repeatable, audit-friendly severity quantification by region (cervical / thoracic / lumbar / SI / hips), enabling registry-grade comparability and longitudinal readiness.

• XR↔MRI complementarity made explicit: a correlation matrix shows how XR carries major phenotype information in this case, while MRI contributes level-specific stenosis / neural compression detail—without changing the primary pattern classification.

• Reproducibility by design: deterministic rendering + descriptor completeness + fixed-order outputs—built for governance, traceability, and downstream reuse.

Who should care (practical value)
Clinicians: A concrete demonstration that a well-captured XR series can carry high phenotype signal—often enough to anchor “mechanical vs inflammatory driver” decisions—while MRI adds targeted anatomic detail rather than replacing the phenotype story.
Pharma/CRO: Cleaner cohort stratification and negative-control framing reduce misclassification, screen failures, and endpoint drift—especially when labels like “AS?” contaminate real-world datasets.
AI/analytics teams: A purposely challenging axial case for stress-testing segmentation, pattern classification, and explainability modules in the presence of artifacts and vascular calcification.

Two appendices

• Appendix A–F: Supplementary tables, matrices, and correlation artifacts supporting the main dossier.

• Experimental Analytics Appendix: Research-only demonstration of additional analytics in a non-enabling format (no proprietary codes/tokens/thresholds, no implementation specifics). It explicitly represents only a subset of platform capability.

Positioning / safety
This publication is an operational demonstration of governed imaging reporting and deterministic analytics. It is not a diagnostic performance claim, not evidence of therapeutic efficacy, and not intended to guide individual clinical care or treatment decisions.

Record contents

• Main case dossier (PDF): narrative + key tables + imaging plates + XR↔MRI correlation

• Appendix A–F 

• Experimental analytics appendix