Inflammatory & Autoimmune Arthritis

The domain that started it all — and the one in which the platform’s architectural commitments are most fully expressed.

Rheumatoid arthritis, psoriatic arthritis, axial and peripheral spondyloarthritis — analyzed through structured longitudinal tracking, phenotype-level differentiation, and treatment-response correlation across timepoints.

RheumaView™ does not score joints in isolation. It builds a governed structural narrative across timepoints — separating true inflammatory progression from degenerative drift, harmonizing reads across centers and across modalities, and producing research-grade endpoints suitable for clinical trials.

The four nosologies covered in this domain share a structural vocabulary, but each carries a distinct phenotype signature, a distinct longitudinal behavior, and a distinct relationship between imaging and treatment response. Each is addressed below with the depth that pharmaceutical sponsors, CROs, academic collaborators, and qualified investors require to evaluate fit.

Inflammatory disease is where the cost of unstructured reading is highest — and where the architectural payoff of a governed pathway is most visible.

Why inflammatory disease, and why first.

Inflammatory and autoimmune arthritis sits at the intersection of three problems that conventional radiographic reading has never fully resolved.

The first is longitudinal coherence. The clinical question in rheumatoid arthritis or spondyloarthritis is rarely «what is on this image» — it is «what has changed since the last image, and is the change inflammatory, degenerative, post-treatment, or projection-driven». Conventional reads answer that question narratively, with phrasing that varies between readers and between timepoints. The variability is not a measurement error. It is a structural property of the workflow.

The second is phenotype heterogeneity. Rheumatoid arthritis, psoriatic arthritis, axial spondyloarthritis, and peripheral spondyloarthritis are not single diseases — each is a family of phenotypes with distinct structural signatures, distinct longitudinal behaviors, and distinct relationships to treatment. A scoring system optimized for one nosology often blurs the others. A scoring system that tries to cover all four loses the discriminative resolution that pharmaceutical trials and translational research require.

The third is inflammatory–degenerative separation. The most consequential diagnostic gray zone in musculoskeletal imaging is not whether disease is present — it is what kind of disease is driving the structural change. Joint-space narrowing, erosive change, and subchondral signal can be produced by inflammatory, degenerative, or mechanical processes, and the three lead to entirely different therapeutic decisions. A read that does not separate them defensibly is a read that travels poorly across centers and across trials.

RheumaView™ began with inflammatory disease because that is where the architectural commitments — deterministic structuring, descriptor lineage, governed comparability — produce the most visible difference from narrative reading.

The platform’s response to these three problems is architectural, not algorithmic. Longitudinal coherence is enforced by a validator-governed pathway that treats every timepoint as a position in a structured chain rather than as an isolated read. Phenotype heterogeneity is preserved by descriptor-level differentiation that does not collapse distinct nosologies into a single composite. Inflammatory–degenerative separation is built into the structural grammar of the system itself — not added downstream as a correction.

The four nosologies described in the sections that follow each illustrate a different aspect of this architecture. Rheumatoid arthritis demonstrates longitudinal tracking. Psoriatic arthritis demonstrates phenotype-level differentiation. Axial spondyloarthritis demonstrates governed comparability across structurally heterogeneous regions. Peripheral spondyloarthritis demonstrates the boundary cases — the patterns that defeat conventional scoring and that a governed pathway is specifically designed to absorb.

Rheumatoid arthritis.

The nosology in which longitudinal tracking is most consequential — and most difficult to do reproducibly.

Rheumatoid arthritis is the canonical longitudinal disease. Its clinical management is governed less by any single image than by the structural trajectory across timepoints — the slope of progression, the response to therapy, the emergence or arrest of erosive change, the differentiation of true inflammatory progression from positional, projection-driven, or degenerative drift.

Conventional scoring systems compress this trajectory into composite indices that travel poorly between readers, between centers, and between trials. The compression is not a flaw of the readers — it is a property of the workflow. A score that summarizes joint-level findings into a single number cannot, by construction, preserve the descriptor-level lineage needed for treatment-response inference or for cohort stratification in pharmaceutical research.

RheumaView™’s response is structural rather than computational. Each joint, each region, each finding is preserved as a discrete descriptor with explicit lineage — traceable from the input image through the validator-governed pathway to the structured clinical output. Across timepoints, the platform does not recompute a score; it constructs a governed comparison in which each descriptor’s behavior is tracked individually, with explicit handling of the projection, positional, and acquisition variables that conventionally introduce uncontrolled noise.

In rheumatoid arthritis, the question is rarely what the image shows. It is what has changed since the last image — and whether the change is the disease, the treatment, the projection, or the patient.

What the domain delivers

For pharmaceutical sponsors and CROs, the platform produces trial-compatible structural endpoints with descriptor-level lineage — outputs that survive the variability problem at its architectural root rather than as a downstream calibration step. Longitudinal slopes, descriptor-level progression signals, and treatment-response correlates are extracted within the same validator-governed pathway that produces the clinician-facing read, with the clinical and research layers held in protected separation.

For academic and translational research, the platform offers descriptor-level export with full lineage — a structured data source from which biostatisticians and translational teams can construct hypotheses, stratify cohorts, and interrogate phenotype-treatment relationships at a resolution that joint-count composites do not preserve.

For health systems and qualified investors, the architectural value is the reproducibility itself. Two reads of the same study, by the same operator or across operators, produce the same structured output — not because the system is calibrated to do so, but because the validator-governed pathway is deterministic by construction. The same property that makes the platform trial-compatible makes it defensible as a longitudinal record across years of clinical follow-up.

The disclosure boundary in this nosology

What is described publicly: the categories of structural finding the platform tracks, the existence of descriptor-level lineage, the architectural commitment to deterministic structuring across timepoints, and the output formats available to clinical and research consumers. What remains proprietary: the validator-chain composition, the descriptor-level rules that enforce inflammatory–degenerative separation, the threshold logic, and the operator-level mechanics of the longitudinal comparison itself.

The public surface is sufficient for fit evaluation. The proprietary layer is what makes the architecture defensible — and what is captured under the patent-pending positioning that governs all RheumaView™ disclosure.

Psoriatic arthritis.

The nosology in which phenotype-level differentiation is the architecture, and composite scoring is the failure mode.

Psoriatic arthritis is not a single disease. It is a family of phenotypes — distal-predominant, oligoarticular asymmetric, polyarticular RA-like, axial-predominant, and arthritis mutilans — each with a distinct structural signature, a distinct longitudinal behavior, and a distinct relationship to therapy.

Compounding the heterogeneity, the disease layers structural lesion types that do not coexist in conventional rheumatoid analysis: enthesitis, dactylitis, periostitis, juxta-articular bone proliferation, central-axial involvement that overlaps with spondyloarthritis, and the destructive–proliferative paradox in which erosive change and new bone formation appear in the same hand at the same timepoint.

A scoring system optimized for any single phenotype loses the others. A scoring system that tries to cover all phenotypes loses the discriminative resolution that pharmaceutical trials and translational research require. The variability is not a measurement error — it is what happens when a heterogeneous family of phenotypes is forced through a homogeneous scoring pipeline.

The architectural answer to phenotype heterogeneity is not a better composite score. It is a refusal to compose.

RheumaView™ preserves phenotype heterogeneity by descriptor-level differentiation. Erosive change, bone proliferation, periosteal reaction, distal involvement, axial involvement, and the soft-tissue correlates of dactylitis and enthesitis are tracked as discrete descriptors with explicit lineage, not collapsed into a single number. Phenotype assignment, where it is made, is constructed from the descriptor pattern — not imposed on it.

The destructive–proliferative paradox, which conventionally produces unstable scoring, is absorbed at the architectural level. Erosive descriptors and proliferative descriptors are distinct objects in the structured output. They can coexist, evolve independently across timepoints, and contribute to phenotype assignment and treatment-response inference without colliding inside a composite.

What the domain delivers

For pharmaceutical sponsors and CROs, the platform produces phenotype-stratified cohorts with descriptor-level inclusion and exclusion criteria — the resolution required for trials targeting specific PsA phenotypes, for label-expansion studies, and for treatment-response analyses where phenotype confounds the outcome. The same descriptor lineage that supports cohort construction supports endpoint extraction within a single validator-governed pathway.

For academic and translational research, the platform offers descriptor-level export across the full PsA spectrum — including the soft-tissue and entheseal correlates that are typically lost in joint-count systems. Phenotype boundaries, phenotype transitions across timepoints, and the relationship between peripheral and axial involvement become tractable as quantitative questions rather than narrative ones.

For health systems and qualified investors, PsA is the nosology in which the architectural payoff of governed, descriptor-level structuring is most strategically visible. The therapeutic landscape is heterogeneous and expanding; the imaging questions that distinguish patient populations are descriptor-level, not score-level; and the data infrastructure required to support phenotype-aware decision-making is precisely what the platform provides.

The disclosure boundary in this nosology

What is described publicly: the families of phenotype the platform differentiates, the categories of structural lesion preserved as discrete descriptors, the architectural commitment to phenotype-level rather than composite-level resolution, and the output formats available to clinical and research consumers. What remains proprietary: the descriptor-level rules that govern phenotype assignment, the validator logic that absorbs the destructive–proliferative paradox, the threshold structure, and the operator-level mechanics of the differentiation itself.

The public surface is sufficient for fit evaluation. The proprietary layer is what makes phenotype-level differentiation defensible as a deterministic property of the architecture — and what is captured under the patent-pending positioning that governs all RheumaView™ disclosure.

Axial spondyloarthritis.

The nosology in which governed comparability is the central architectural achievement — across regions, modalities, and decades.

Axial spondyloarthritis poses a structural problem that no other inflammatory nosology poses with the same intensity. Disease activity and disease damage are read across structurally heterogeneous regions — sacroiliac joints, vertebral bodies, posterior elements, ligamentous insertions, costovertebral and costotransverse articulations — each with its own anatomical grammar, its own projection sensitivity, and its own characteristic lesion vocabulary.

The disease unfolds across timescales that exceed any single trial window. Erosive sacroiliitis, sclerotic remodeling, syndesmophyte formation, ankylotic bridging, and the late-stage architectural changes of advanced disease can take years or decades to evolve. Reads acquired across that interval are reads acquired by different operators, on different equipment, with different acquisition protocols — and yet the clinical and research questions require that they be comparable.

Conventional scoring systems for axial disease are explicit about this difficulty. They compress findings into composite indices that have known reader-dependent variance, known projection sensitivity, and known limitations in distinguishing nascent inflammatory change from established structural damage. The variance is documented; what has not historically existed is an architectural alternative.

The axSpA problem is not better measurement. It is governed comparability — across regions that do not share a grammar, across modalities that do not share a substrate, and across timepoints that do not share an operator.

RheumaView™ addresses axial disease through region-specific descriptor frameworks held within a single validator-governed pathway. Sacroiliac findings, vertebral findings, ligamentous findings, and the cross-regional patterns that distinguish axSpA from mimics are tracked as discrete descriptors with explicit lineage. Region-specific projection and acquisition variables are handled within the pathway, not as downstream corrections — which is what makes longitudinal comparison defensible across the heterogeneous reading conditions that axial disease invariably accumulates.

Cross-modality concordance — the relationship between plain radiographic findings and MRI correlates, where both are available — is preserved at descriptor level rather than collapsed into a single read. The same architectural property allows the platform to operate downstream of upstream imaging AI when such systems are already deployed: the validator-governed pathway applies its deterministic structuring, descriptor lineage, and protected clinical–research separation regardless of upstream source.

What the domain delivers

For pharmaceutical sponsors and CROs, the platform produces harmonized multi-site endpoints for axSpA trials — the use case in which cross-reader and cross-center variance has been the most persistent obstacle to defensible structural endpoints. Descriptor-level export supports modified-trial-design questions that joint-region composites do not preserve: nascent inflammatory change versus established damage, region-specific progression rates, and treatment-response signals that vary by anatomical compartment.

For academic and translational research, the platform offers descriptor-level export across the full axial domain, with cross-modality concordance preserved where multi-modal acquisition is available. Long-arc longitudinal questions — the natural history of structural progression, the divergence of imaging trajectories from clinical activity, the relationship between early inflammatory signal and late architectural damage — become tractable as quantitative inquiries with descriptor-level resolution.

For health systems and qualified investors, axSpA is the nosology in which the architectural cost of unstructured reading compounds most visibly across the patient lifetime. The platform’s reproducibility property — that two reads of the same study, by the same or different operators, produce the same structured output — is what makes the longitudinal record defensible across the decade-scale arc on which clinical and trial decisions in axial disease ultimately depend.

The disclosure boundary in this nosology

What is described publicly: the regions and lesion categories the platform tracks, the architectural commitment to region-specific descriptor frameworks within a single governed pathway, the principle of cross-modality concordance preservation, and the output formats available to clinical and research consumers. What remains proprietary: the validator logic that enforces region-specific projection handling, the descriptor-level rules that govern syndesmophyte and ankylotic progression tracking, the cross-modality concordance mechanics, and the threshold structure underlying long-arc longitudinal comparison.

The public surface is sufficient for fit evaluation. The proprietary layer is what makes governed comparability defensible across heterogeneous regions, modalities, and decades — and what is captured under the patent-pending positioning that governs all RheumaView™ disclosure.

Peripheral spondyloarthritis.

The nosology in which the boundary cases — the patterns conventional pipelines push to the margin — return as first-class structural data.

Peripheral spondyloarthritis is the nosology that defeats template-based reading. Its phenotypes — reactive arthritis, IBD-associated arthritis, undifferentiated peripheral spondyloarthritis, and the peripheral-predominant presentations of psoriatic and axial disease — share an entheseal architecture, a soft-tissue inflammatory substrate, and a pattern of structural change that does not reduce to the joint-count grammar built for rheumatoid disease.

Enthesitis, dactylitis, asymmetric oligoarticular involvement, lower-limb predominance, periosteal reaction, juxta-articular bone proliferation, and the entheseal-bone interface that distinguishes spondyloarthritis from rheumatoid pathology are the structural signatures of peripheral disease. Each is anatomically and pathologically real, each is documented in the literature, and each is poorly served by scoring systems that operate at joint level and treat soft-tissue and entheseal correlates as out-of-scope.

The result, in conventional pipelines, is that the structural signal of peripheral disease lives in the radiologist’s narrative — visible, clinically meaningful, but unreachable as data. It does not survive the transition from read to record. It does not survive aggregation across centers. It does not survive trial endpoint extraction. It is consistently the first thing lost when imaging is asked to behave as structured information.

The test of a governed pathway is not whether it handles the canonical cases. It is whether it absorbs the boundary cases — the entheseal, the asymmetric, the soft-tissue — without pushing them out of the structured record.

RheumaView™ treats the entheseal–soft-tissue–periosteal substrate as first-class descriptor territory rather than as out-of-scope narrative. Enthesitis, dactylitis, periostitis, juxta-articular proliferation, and the entheseal-bone interface findings that define peripheral spondyloarthritis are tracked as discrete descriptors with explicit lineage, within the same validator-governed pathway that produces the joint-level structural read. Asymmetric oligoarticular patterns are preserved as patterns — not normalized into a symmetric template that erases the diagnostic signal.

The architectural property at work here is the same one that operates throughout the inflammatory domain — descriptor-level resolution rather than composite collapse — but its consequences in peripheral spondyloarthritis are most visible. The findings that conventional pipelines lose are precisely the findings that distinguish peripheral spondyloarthritis from its mimics, that stratify cohorts in trials, and that mark the structural signal of treatment response in entheseal-driven disease.

What the domain delivers

For pharmaceutical sponsors and CROs, the platform produces descriptor-level cohort construction for peripheral spondyloarthritis trials — including the entheseal, dactylitic, and periosteal findings on which inclusion criteria, stratification, and response endpoints in this disease ultimately depend. Treatment-response inference is preserved at the level of the lesions that actually carry the signal, not collapsed into joint-count composites that were designed for a different pathology.

For academic and translational research, the platform offers descriptor-level export across the full peripheral spondyloarthritis spectrum, including the entheseal–soft-tissue–periosteal correlates that conventional pipelines lose. Boundary questions — the differentiation of peripheral spondyloarthritis from peripheral-predominant psoriatic disease, from rheumatoid disease with atypical distribution, from reactive and post-infectious arthritides — become tractable as quantitative inquiries with descriptor-level lineage rather than as narrative judgments.

For health systems and qualified investors, peripheral spondyloarthritis is the nosology in which the architectural commitment to descriptor-level resolution earns its strongest validation. A platform that preserves the boundary cases preserves the canonical cases by construction. A platform that loses them — that pushes entheseal and soft-tissue findings out of the structured record — is a platform whose architectural commitments do not extend to where the clinical and trial questions are actually most demanding.

The disclosure boundary in this nosology

What is described publicly: the categories of entheseal, dactylitic, periosteal, and asymmetric structural finding the platform preserves as discrete descriptors, the architectural commitment to descriptor-level rather than composite-level resolution, and the output formats available to clinical and research consumers. What remains proprietary: the descriptor-level rules that govern entheseal and soft-tissue assignment, the validator logic that handles asymmetric and oligoarticular patterns, the threshold structure, and the operator-level mechanics underlying boundary-case absorption.

The public surface is sufficient for fit evaluation. The proprietary layer is what makes boundary-case absorption defensible as a deterministic property of the architecture — and what is captured under the patent-pending positioning that governs all RheumaView™ disclosure.

The longitudinal architecture, in one frame.

What the four nosologies share is not a scoring system. It is a pathway.

Across rheumatoid arthritis, psoriatic arthritis, axial and peripheral spondyloarthritis, the platform’s response to longitudinal reading is the same architecture — applied with nosology-specific descriptor frameworks, but governed by a single deterministic pathway.

Each timepoint enters the pathway as an independent input. The pathway preserves descriptor lineage from input to output. Across timepoints, the platform constructs a governed comparison in which descriptors are tracked individually rather than re-aggregated into a composite. Two outputs emerge from a single pass: the clinician-facing structured read and the research-facing analytic addendum. The two are produced from the same data, by the same pathway, under protected separation.

The architectural commitment is not faster reading or better scoring. It is that the same study, read on the same pathway, produces the same structured output — and that this property holds across timepoints, across centers, and across the four nosologies of the inflammatory domain.

What descriptor-level lineage gives a sponsor.

A composite tells you the score moved. A descriptor tells you what moved, where, and in which direction.

Treatment-response inference in inflammatory arthritis trials is constrained by the resolution of the imaging endpoint. A composite score that has moved by one or two units across a treatment interval cannot, by construction, distinguish between regression of inflammatory change, arrest of progression, post-treatment remodeling, and reader drift. The four explanations have different therapeutic implications. The composite cannot separate them.

The platform’s descriptor-level output is designed for the question one resolution finer. Each descriptor — joint-level erosive change, region-specific bone proliferation, entheseal involvement, soft-tissue correlate, axial-region structural finding — carries its own trajectory, its own lineage, and its own behavior across the treatment interval. A treatment effect that operates on entheseal disease but not on erosive disease, or on axial findings but not on peripheral, is visible as a pattern across descriptors rather than as a number that has or has not moved.

For trial endpoints, this resolution is the difference between a defensible mechanism-of-action signal and a composite that confounds it. For label-expansion studies, it is the difference between identifying the phenotype that responds and recruiting a heterogeneous cohort that dilutes the signal. For regulatory submissions, it is the difference between a structural endpoint with descriptor-level lineage and one that depends on reader-dependent composite assignment.

The clinical question in a trial is rarely whether the score moved. It is which descriptors moved, in which direction, in which subpopulation — and whether that pattern is consistent with the treatment’s mechanism of action.

What is operationally available

The platform produces, from a single pass through the validator-governed pathway, a clinician-facing structured read and a research-facing analytic addendum under protected separation. The research layer carries descriptor-level lineage with traceable provenance from input image through structured output — the lineage requirement that defensible regulatory and translational endpoints depend on.

For multi-center trials, the same architectural property — deterministic structuring within a single governed pathway — produces harmonized output across sites, across operators, and across acquisition protocols, without requiring downstream calibration as a remediation step. Cross-reader and cross-center variance is addressed at the architectural root rather than as a recovery operation against a workflow that produced it.

Where upstream imaging AI is already deployed in a sponsor’s infrastructure, the validator-governed pathway operates downstream of those outputs and applies the same deterministic structuring, descriptor lineage, and protected clinical–research separation regardless of upstream source. The architecture is not a replacement for existing detection or analytic tools; it is a layer that governs them.

What is excluded by design

The platform does not assign treatment recommendations. It does not predict treatment response. It does not generate clinical decisions. It produces structured imaging output with descriptor-level lineage — the substrate from which sponsors, clinicians, and translational teams construct decisions, hypotheses, and endpoints within their own analytical and regulatory frameworks.

The architectural restraint is not incidental. A platform that crosses from structured imaging into clinical or therapeutic recommendation acquires regulatory exposure, reader-dependent variance at a different layer, and a defensibility surface that is harder to govern. RheumaView™ stops at the structured imaging boundary by design — and the proprietary layer that makes the boundary defensible is captured under the patent-pending positioning that governs all platform disclosure.

What descriptor-level lineage gives a sponsor.

A composite tells you the score moved. A descriptor tells you what moved, where, and in which direction.

Treatment-response inference in inflammatory arthritis trials is constrained by the resolution of the imaging endpoint. A composite score that has moved by one or two units across a treatment interval cannot, by construction, distinguish between regression of inflammatory change, arrest of progression, post-treatment remodeling, and reader drift. The four explanations have different therapeutic implications. The composite cannot separate them.

The platform’s descriptor-level output is designed for the question one resolution finer. Each descriptor — joint-level erosive change, region-specific bone proliferation, entheseal involvement, soft-tissue correlate, axial-region structural finding — carries its own trajectory, its own lineage, and its own behavior across the treatment interval. A treatment effect that operates on entheseal disease but not on erosive disease, or on axial findings but not on peripheral, is visible as a pattern across descriptors rather than as a number that has or has not moved.

For trial endpoints, this resolution is the difference between a defensible mechanism-of-action signal and a composite that confounds it. For label-expansion studies, it is the difference between identifying the phenotype that responds and recruiting a heterogeneous cohort that dilutes the signal. For regulatory submissions, it is the difference between a structural endpoint with descriptor-level lineage and one that depends on reader-dependent composite assignment.

The clinical question in a trial is rarely whether the score moved. It is which descriptors moved, in which direction, in which subpopulation — and whether that pattern is consistent with the treatment’s mechanism of action.

What is operationally available

The platform produces, from a single pass through the validator-governed pathway, a clinician-facing structured read and a research-facing analytic addendum under protected separation. The research layer carries descriptor-level lineage with traceable provenance from input image through structured output — the lineage requirement that defensible regulatory and translational endpoints depend on.

For multi-center trials, the same architectural property — deterministic structuring within a single governed pathway — produces harmonized output across sites, across operators, and across acquisition protocols, without requiring downstream calibration as a remediation step. Cross-reader and cross-center variance is addressed at the architectural root rather than as a recovery operation against a workflow that produced it.

Where upstream imaging AI is already deployed in a sponsor’s infrastructure, the validator-governed pathway operates downstream of those outputs and applies the same deterministic structuring, descriptor lineage, and protected clinical–research separation regardless of upstream source. The architecture is not a replacement for existing detection or analytic tools; it is a layer that governs them.

What is excluded by design

The platform does not assign treatment recommendations. It does not predict treatment response. It does not generate clinical decisions. It produces structured imaging output with descriptor-level lineage — the substrate from which sponsors, clinicians, and translational teams construct decisions, hypotheses, and endpoints within their own analytical and regulatory frameworks.

The architectural restraint is not incidental. A platform that crosses from structured imaging into clinical or therapeutic recommendation acquires regulatory exposure, reader-dependent variance at a different layer, and a defensibility surface that is harder to govern. RheumaView™ stops at the structured imaging boundary by design — and the proprietary layer that makes the boundary defensible is captured under the patent-pending positioning that governs all platform disclosure.

What you have read is the surface.

The four nosologies, the longitudinal architecture, the descriptor-level treatment-response inference, and the three engagement channels constitute the public face of the inflammatory and autoimmune arthritis domain.

What is described publicly: the categories of structural finding the platform tracks; the architectural commitments to deterministic structuring, descriptor lineage, and governed comparability; the principle of clinical–research separation produced from a single pass; the disease boundaries the domain covers; and the output formats available to clinical and research consumers.

What remains proprietary: the validator-chain composition and stage logic; the descriptor-level rules that govern phenotype assignment and inflammatory–degenerative separation; the threshold structure underlying longitudinal comparison; the operator-level mechanics of cross-modality concordance and boundary-case absorption; and the internal architecture of the protected separation between clinical and research layers.

RheumaView™ is intended for use by licensed medical professionals and qualified research environments. Not a patient-facing diagnostic tool. Public materials describe categories, architecture, and domain breadth; implementation details — validator rules, descriptor formulas, operator mechanics, threshold structure — remain proprietary.