1.1 The early seronegative diagnostic gap

Early inflammatory arthritis does not uniformly present with a clear external phenotype. In seronegative patients — those lacking RF and anti-CCP positivity — clinical features may be intermittent or low-amplitude, and the initial differential frequently includes mechanical, crystal-associated, and degenerative pathways. The result is a predictable uncertainty gap: true inflammatory disease may be undertreated (diagnostic inertia), while benign mimics may be overtreated when ambiguous imaging findings are overinterpreted. Both failure modes carry cost during the period when disease modification is most impactful.

In this context, imaging is often treated as a confirmatory step, yet in practice it can either accelerate clarity or inadvertently reinforce uncertainty. A radiograph read as “no erosions” may be technically correct but clinically non-actionable, because the decision question in early seronegative disease is rarely binary. The clinically decisive questions are typically:

• Is the dominant driver at the symptomatic sites inflammatory or non-inflammatory (and where is it — joint, tendon sheath, enthesis)?
• Does the distribution and morphology support an RA-like phenotype, a PsA/SpA-like phenotype, or a mimic pathway?
• Is there imaging context suggesting higher structural risk, and what is the next best test to reduce uncertainty?

1.2 Why first-line imaging often fails

First-line imaging fails early seronegative patients less because information is absent and more because it is inconsistently sought and described. Routine narrative reports may omit distribution statements, joint-by-joint mapping, tendon sheath and enthesis assessment, and explicit mimic analysis. When subtle cortical abnormalities are present, they may be either undercalled (“normal”) or overcalled (“early erosions”) without the pattern context needed for safe interpretation.

This review frames the gap as an actionability problem: the clinician needs a structured description that enables probability updates and targeted escalation, not a generic normal/abnormal label.

1.3 Imaging-first phenotype anchoring: definition and intent

We use the term imaging-first phenotype anchoring to describe a probability-shifting workflow in which imaging patterns are used to strengthen or weaken competing phenotypes in early seronegative disease. The intent is not diagnosis by image alone, nor universal escalation to advanced imaging, but a disciplined approach to (i) detect subtle structural and periarticular patterns, (ii) explicitly manage mimics, and (iii) deploy US-PD or MRI selectively when they answer a discriminative clinical question.

Within peripheral disease, two early pattern elements are emphasized: microerosions (small cortical breaks or margin defects that become meaningful in anatomical and distribution context) and enthesitis-adjacent changes (peri-entheseal cortical irregularity and related features that can anchor PsA/SpA-like pathways when paired with activity context).

1.4 Scope statement

This review focuses on peripheral joints and periarticular structures — hands, wrists, feet, ankles, and selected elbows/knees when clinically relevant. Early axial skeleton and pelvic/hip pattern recognition involves distinct lesion taxonomy and modality priorities and will be addressed in a companion manuscript.

2.1 Defining “early” and “seronegative”

For practical purposes, early refers to patients within approximately the first 12–24 months of persistent symptoms or documented inflammatory episodes, recognizing that onset is often imprecise. Seronegative refers to absence of RF and anti-CCP positivity; additional markers (e.g., HLA-B27) are treated as contextual rather than defining. The framework is designed for situations where serology does not provide a phenotype anchor and imaging is asked to reduce diagnostic uncertainty.

2.2 The “low external phenotype” presentation

The core use case is the patient with inflammatory-suggestive pain — morning stiffness, episodic swelling, inflammatory flares, or objective tenderness patterns — but limited external phenotype: minimal swelling on exam, absent deformity, and sometimes absent or occult psoriasis. In such cases, periarticular drivers (tenosynovitis, enthesitis) may dominate symptoms while conventional radiographs remain near-normal.

2.3 Decision stakes — delay, mismatch, progression

Decision stakes are asymmetric. Under-recognition of inflammatory disease can delay therapy intensification and allow structural progression; over-interpretation of ambiguous microfindings can lead to unnecessary immunomodulation. Imaging-first phenotype anchoring is intended to reduce both errors by requiring (i) explicit pattern context, (ii) mimic controls, and (iii) question-driven escalation to US-PD or MRI only when the result is likely to change management posture or follow-up intensity.

This narrative review was informed by targeted reading of guideline documents and primary studies addressing imaging in early inflammatory arthritis, seronegative rheumatoid arthritis, prognostic value of imaging lesions, and imaging mimics. The goal is not exhaustive synthesis but an evidence-anchored, implementation-oriented framework for peripheral early seronegative presentations.

3.1 Professional guidelines

Current guidelines position imaging as an adjunct when clinical doubt persists, with modality selection driven by the question being asked — structural baseline, active inflammation, or alternative diagnoses.^[1–4] Recommendations recognize that radiography provides a structural baseline and distribution context, ultrasound can detect synovitis/tenosynovitis and Doppler activity, and MRI can identify osteitis/BME and occult inflammation when it is decision-relevant.

3.2 Imaging-guided treatment trials: ARCTIC and IMAGINE-RA

Two RCTs are frequently cited. ARCTIC^[5] compared ultrasound-guided tight control with conventional tight control in early RA and did not demonstrate superior outcomes with routine ultrasound-driven escalation. IMAGINE-RA^[6,7], focused on patients in clinical remission, similarly did not show benefit of an MRI-guided treat-to-target strategy compared with conventional treat-to-target care.

These trials evaluate routine imaging targets as escalation triggers in established or controlled disease. They do not negate the role of imaging for early phenotype anchoring in seronegative uncertainty.

3.3 Ultrasound in seronegative RA

Multiple studies support the value of musculoskeletal ultrasound — particularly grayscale synovial hypertrophy, PD activity, tenosynovitis, and ultrasound-detected erosions — in differentiating seronegative RA from OA and other causes of symptoms.^[8,9] This literature reinforces a central practical point: inflammatory activity may be present and localizable even when radiographs are non-specific.

3.4 Prognostic value of imaging findings

Observational studies in early arthritis suggest that combinations of structural lesions (including small erosive changes) and objective activity (Doppler signal on ultrasound) are associated with higher likelihood of subsequent structural progression.^[10,11] In MRI literature, osteitis/BME has repeatedly been linked to increased risk of radiographic progression and erosive development in early disease.^[12–14]

3.5 Advanced imaging modalities

HR-pQCT can characterize cortical breaks and microstructural bone change at a resolution unavailable to routine clinical modalities and is often used as a research reference for microerosion concepts.^[15] Its role here is primarily conceptual — supporting the plausibility that microstructural cortical change exists early — rather than as a routine clinical recommendation.

3.6 Mimics and pseudoerosions

A safety-critical body of work highlights pseudoerosions and other mimics — anatomic concavities, vascular channels, degenerative pits, subchondral cysts, projection artifacts — that can be mistaken for inflammatory erosions.^[16] This literature emphasizes that “micro-defects” require multi-view confirmation, distribution logic, and correlation with co-features and activity context.

Crystal-associated arthropathies represent a distinct mimic category.^[4] In patients where gout or CPPD enters the differential, DECT provides definitive detection of MSU crystal deposits and may clarify whether erosive changes reflect crystal-driven versus inflammatory mechanisms. Ultrasound can also detect crystal deposits (double contour sign, hyperechoic aggregates) but with lower specificity than DECT.

4.1 Plain radiography (XR)

Radiography remains the first-line modality for many suspected inflammatory arthritis presentations because it provides a standardized structural baseline, broad joint-set coverage, and efficient pattern screening.^[1,3] XR is particularly valuable for (i) mapping distribution and symmetry across hands/feet, (ii) identifying alignment and biomechanics that may drive symptoms or mimic inflammation, and (iii) establishing longitudinal comparability for detecting change over time.

Limitation. XR is less sensitive for early inflammatory activity and may miss marrow and soft-tissue disease. An XR report stating “no erosions” should be treated as a baseline statement, not as exclusion of inflammatory pathology when clinical suspicion persists.

Scenarios where XR can be superior

Radiographs may outperform US and, in focused scenarios, MRI for mineralized cortical pathology and standardized structural comparison. XR can be superior for confirming cortical integrity of suspected microerosion candidates across complementary projections, detecting early mineralized periosteal response and entheseal new bone, assessing joint space narrowing and alignment mechanics with reproducible geometry, and recognizing whole-hand/whole-foot distribution patterns in a single exam. By contrast, ultrasound is operator- and window-dependent for cortical surfaces and cannot assess intraosseous processes; MRI excels for BME and soft tissue inflammation but is less efficient for whole-joint-set pattern mapping.

4.2 Ultrasound with power Doppler (US-PD)

US-PD is a high-yield second-line modality when the decisive question is whether symptoms map to active inflammation and where that activity resides — synovium, tendon sheath, or enthesis.^[8–11] Strengths include real-time assessment, detection of synovitis and tenosynovitis, and Doppler activity as an objective inflammatory marker. Limitations include operator dependence, variability in scanning protocols, and restricted ability to survey the entire joint set at high detail in routine practice.

4.3 Magnetic resonance imaging (MRI)

MRI is most valuable when it answers a discriminative question that XR and US-PD cannot resolve: presence and extent of osteitis/BME as a risk modifier, occult synovitis/tenosynovitis in complex regions, or deep periarticular inflammation.^[12–14] MRI should be used selectively in an XR-first pathway — particularly when discordance persists or when marrow-level inflammatory context would change follow-up intensity or therapy posture.

4.4 Modality utility matrix

A pragmatic modality utility matrix summarizes where each modality is most likely to change decisions in early seronegative peripheral disease. Utility is question-dependent and assumes adequate technique.

Table 1. Modality utility matrix for early seronegative peripheral inflammatory arthritis. Legend: +++ high decision impact; ++ moderate; + supportive; − usually not decision-changing (assuming adequate technique). For crystal deposition, DECT provides definitive MSU detection when gout is in the differential; US can detect double contour sign and hyperechoic aggregates with lower specificity.

5.1 Defining microerosions in early disease

Because “microerosion” is used variably in clinical discourse, this review adopts an operational definition: a microerosion candidate is a focal cortical break or margin defect that is anatomically plausible for inflammatory involvement and is supported by

• reproducibility across projections or consistent localization, and/or
• coherent distribution and co-features — synovitis/tenosynovitis, peri-entheseal change, or MRI inflammatory context.

The label “candidate” is intentional and signals the need for pattern context and mimic control rather than implying diagnostic certainty.

5.2 Mimics and pseudo-erosions

Microerosion candidates are most commonly confounded by pseudoerosions and degenerative or anatomic variants.^[16] Key pitfalls include vascular channels, subchondral cysts/geodes, degenerative pits, traction-related cortical irregularity, and projectional pseudo-defects. Operational checkpoints for practice include

• multi-view confirmation;
• surface-specific description (marginal vs central);
• distribution logic (which joints, symmetry); and
• correlation with activity context (US-PD) when available.

5.3 Enthesitis-adjacent changes

Enthesitis-adjacent changes refer here to peri-entheseal cortical irregularity and related mineralized responses that gain phenotype value when paired with distribution patterns and activity context. In early seronegative disease, these features can shift probability toward PsA/SpA-like pathways — particularly when accompanied by tendon/enthesis inflammation on US-PD or by regional inflammatory context on MRI — while recognizing that enthesis-adjacent mineralization can also be degenerative or mechanical.

Peripheral showcase regions for early pattern recognition include MCP2–3 margins, ulnar styloid/carpus, and MTP joints (particularly MTP5), supplemented by PIP/DIP depending on the phenotype hypothesis. Patterns below function as probability updates and end with a “next best test” question.

Pattern A — marginal microerosion candidates at classic RA sites

Description. Small marginal cortical defects at classic RA-prone sites — MCP margins, ulnar styloid/carpal interfaces, MTP margins — that cluster in a symmetric or RA-leaning distribution. Interpretive context strengthens when co-features such as tenosynovitis or synovitis activity are present.

XR emphasis note. Ensure complementary projections that accentuate cortical rims and reduce pseudoerosion risk; describe surface (marginal) and joint-level location explicitly.

Best next test. US-PD targeted to symptomatic joints and tendon compartments to confirm activity and localize drivers; consider MRI if osteitis/BME assessment would change risk posture.

Pattern B — enthesitis-adjacent cortical changes

Description. Peri-entheseal cortical irregularity with distribution and morphology suggestive of an enthesitis-driven pathway, especially when paired with tendon/enthesis activity signals or mild osteoproliferative context.

XR emphasis note. Describe the enthesis-adjacent location and mineralized response (if present) and distinguish from diffuse degenerative enthesopathy by distribution and co-features.

Best next test. US-PD of the relevant entheses and adjacent tendon sheaths; MRI selectively when deep involvement or BME is suspected.

Pattern C — periosteal reaction and fluffy new bone

Description. Periosteal reaction or fluffy new bone in a peripheral distribution that, when integrated with clinical and soft-tissue context, can support PsA/SpA-like phenotypes over purely erosive RA-like pathways.

XR emphasis note. XR is often the most direct modality for thin mineralized periosteal response; describe location and extent and correlate with adjacent soft-tissue drivers when available.

Best next test. US-PD for enthesitis/tenosynovitis activity mapping; MRI if marrow-level inflammatory context is needed.

Pattern D — DIP involvement with nail changes

Description. DIP-predominant involvement — structural change and/or periarticular patterning — in a context where nail findings are present or suspected, shifting phenotype probability toward psoriatic pathways or selected mimic considerations.

XR emphasis note. Carefully map DIP distribution and distinguish erosive OA patterns by osteophyte/sclerosis context and central vs marginal localization.

Best next test. US-PD of DIP synovitis and extensor tendon insertions if activity localization will change posture; consider MRI selectively for complex cases.

Pattern E — asymmetric oligoarticular distribution

Description. Asymmetric oligoarticular distribution with periarticular drivers (tenosynovitis/enthesitis) and limited classic RA symmetry, which can support PsA/SpA-like pathways or reactive/post-infectious phenotypes depending on clinical context.

XR emphasis note. XR can be particularly useful here to document joint-set selection and exclude a predominantly degenerative architecture.

Best next test. US-PD to localize active drivers at the selected joints; MRI only if deep or discordant features remain unresolved.

7.1 The problem of omission bias

In early seronegative disease, report non-actionability is commonly driven by omission bias: absence of joint-by-joint mapping, lack of distribution and symmetry statements, and incomplete assessment of tendon sheaths and entheses. This can create false reassurance (“normal”) or unsafe escalation (“erosions”) when subtle features are not contextualized.

7.2 What rheumatology needs from imaging reports

Actionable reports in this setting should provide:

• joint-level localization (which surface, which joint, laterality);
• distribution and symmetry statements;
• periarticular driver assessment (tenosynovitis, enthesis-adjacent changes);
• mimic checkpoints when microerosion candidates are described; and
• explicit next-test logic when uncertainty remains — US-PD vs MRI and the question to be answered.

7.3 Structured reporting pipelines

Beyond modality choice, diagnostic lift often depends on whether subtle findings are described consistently across readers and timepoints. Template-enforced, descriptor-complete reporting — where required elements (distribution, cortical versus central localization, periarticular drivers, mimics, and uncertainty) are systematically prompted and internally checked for completeness — can reduce variability and improve longitudinal comparability. Such structured workflows function as a reproducibility layer that makes early subtle changes more legible for clinical decision-making without altering the underlying evidence limits of each modality.

8.1 XR-first, question-driven escalation

The proposed escalation framework is XR-first: obtain a standardized structural baseline and distribution map, then escalate to US-PD when the missing information is localization of active inflammation (synovitis/tenosynovitis/enthesitis) or when XR microerosion candidates require activity context. Escalate to MRI selectively when marrow-level inflammatory context (osteitis/BME) or deep anatomy is likely to change risk posture or resolve persistent discordance.

8.2 Order language templates

Order language should force an actionable answer by specifying the phenotype question and the discriminative target.

• For US-PD: request joint and tendon-compartment mapping with semiquantitative severity and PD activity, and specify entheses of concern.
• For MRI: request assessment for synovitis/tenosynovitis and osteitis/BME in the symptomatic region, and ask for distribution/context statements relevant to inflammatory versus degenerative pathways.

9.1 Imaging as probability and risk modifier

In early seronegative disease, imaging should be treated as a modifier of phenotype probability and structural risk, integrated with clinical assessment within treat-to-target principles.^[19] Imaging is not a stand-alone trigger for therapy decisions; rather, it helps determine whether the dominant driver is inflammatory, which phenotype is most plausible, and whether risk context supports closer follow-up and earlier consideration of therapy intensification.

9.2 When imaging may support earlier therapy intensification

Imaging features that increase the pretest probability of persistent inflammatory disease and raise concern for structural progression include coherent inflammatory distribution patterns (RA-like marginal microerosion candidates in classic sites or PsA/SpA-like enthesis-driven patterns), objective activity markers (moderate-to-marked PD signal, tenosynovitis), and/or MRI osteitis/BME in the symptomatic region. In such settings, imaging may justify closer monitoring and may support consideration of earlier therapy intensification when aligned with clinical trajectory and guideline-based treat-to-target care.

9.3 When imaging may support re-triage toward mimics

Conversely, imaging that demonstrates a predominantly degenerative/mechanical architecture (alignment-driven changes, osteophyte/sclerosis-dominant patterns, central erosive OA patterns) without convincing activity context lowers the probability that active inflammatory synovitis is the dominant driver at the imaged sites. These findings support re-triage toward mechanical/crystal pathways or watchful reassessment, with targeted US-PD reserved for cases where activity confirmation would change the plan.

10.1 Reader variability and protocol dependence

Detection of microerosion candidates and enthesis-adjacent change is sensitive to acquisition quality, projection selection, scanning protocol (for US), and reader search behavior. Standardizing views, defining required descriptors, and explicitly documenting projection adequacy reduce variability and improve comparability.

10.2 False positives and false negatives

False positives arise from pseudoerosions and degenerative variants being labeled as inflammatory; false negatives arise from omission bias and sub-threshold disease. A safe framework therefore requires (i) explicit mimic checkpoints, (ii) pattern-context interpretation, and (iii) targeted escalation to US-PD or MRI when uncertainty is decision-relevant.

10.3 Implementation constraints

Access to US-PD and MRI varies, and operator expertise is heterogeneous. The framework is designed to remain scalable by emphasizing XR-first pattern mapping and question-driven escalation rather than universal advanced imaging.

1. Colebatch AN, Edwards CJ, Østergaard M, et al. EULAR recommendations for the use of imaging of the joints in the clinical management of rheumatoid arthritis. Ann Rheum Dis 2013;72(6):804–814.
2. Mandl P, Navarro-Compán V, Terslev L, et al. EULAR recommendations for the use of imaging in the diagnosis and management of spondyloarthritis in clinical practice. Ann Rheum Dis 2015;74(7):1327–1339.
3. Subhas N, Argin M, Engel A, et al. ACR Appropriateness Criteria® Inflammatory Arthritides. J Am Coll Radiol 2023;20(5S):S20–S32.
4. Mandl P, Grunke M, Glinatsi D, et al. EULAR recommendations for the use of imaging in diagnosis and management of crystal-induced arthropathies in clinical practice. Ann Rheum Dis 2024;83:752–759.
5. Haavardsholm EA, Aga AB, Olsen IC, et al. Ultrasound in management of rheumatoid arthritis: ARCTIC randomised controlled strategy trial. BMJ 2016;354:i4205.
6. Møller-Bisgaard S, Hørslev-Petersen K, Ejbjerg B, et al. Effect of MRI vs conventional treat-to-target strategies on disease activity remission and radiographic progression in RA: The IMAGINE-RA randomized clinical trial. JAMA 2019;321(5):461–472.
7. Møller-Bisgaard S, Hørslev-Petersen K, Ørnbjerg LM, et al. Long-term efficacy of a 2-year MRI treat-to-target strategy: 5-year follow-up of the IMAGINE-RA randomised trial. RMD Open 2024;10(1):e003945.
8. Wang H, Liu X, Li Z, et al. The value of high-frequency ultrasound in differentiating seronegative rheumatoid arthritis from osteoarthritis. Sci Rep 2022;12:21372.
9. Cen X, Wang Y, Chen L, et al. Musculoskeletal ultrasound for early diagnosis of seronegative rheumatoid arthritis. J Ultrasound Med 2024;43(5):897–906.
10. Funck-Brentano T, Etchepare F, Gousset J, et al. Benefits of ultrasonography in the management of early arthritis: ESPOIR cohort. Arthritis Care Res 2013;65(6):896–902.
11. Sreerangaiah D, Grayer M, Fisher BA, et al. Quantitative power Doppler US measures predict radiographic progression. Rheumatology 2016;55(1):89–93.
12. Haavardsholm EA, Bøyesen P, Østergaard M, et al. MRI findings in 84 patients with early RA: bone marrow oedema predicts erosive progression. Ann Rheum Dis 2008;67(6):794–800.
13. Hetland ML, Ejbjerg B, Hørslev-Petersen K, et al. MRI bone oedema is the strongest predictor of subsequent radiographic progression (CIMESTRA). Ann Rheum Dis 2009;68(3):384–390.
14. McQueen FM, Ostendorf B. What is MRI bone oedema in rheumatoid arthritis and why does it matter? Arthritis Res Ther 2012;14(4):222.
15. Klose-Jensen R, Tello-Winczek K, Engdahl C, et al. SPECTRA: A large collaborative project for HR-pQCT research in rheumatology. Rheumatology 2020;59(12):3841–3850.
16. Hirtler L, Gmeiner A, Gasser B, et al. Erosions versus pseudoerosions in rheumatoid arthritis: A systematic review. Semin Arthritis Rheum 2019;49(3):365–372.
17. Dale J, Stirling A, Zhang R, et al. Targeting ultrasound remission in early RA: results of the TaSER study. Ann Rheum Dis 2016;75(6):1043–1050.
18. Naredo E, Valor L, De la Torre I, et al. Predictive value of Doppler US-detected synovitis in relation to failed tapering of biologic therapy in RA. Rheumatology 2015;54(8):1408–1414.
19. Smolen JS, Landewé RBM, Bergstra SA, et al. EULAR recommendations for the management of rheumatoid arthritis: 2022 update. Ann Rheum Dis 2023;82(1):3–18.
20. Aletaha D, Neogi T, Silman AJ, et al. 2010 Rheumatoid arthritis classification criteria: ACR/EULAR collaborative initiative. Ann Rheum Dis 2010;69(9):1580–1588.

Acknowledgments

The author acknowledges the contribution of structured reporting frameworks to the concepts presented in this review. A dedicated companion paper will address early axial and pelvic/hip imaging patterns, where lesion taxonomy, biomechanics, and modality priorities differ substantially from peripheral disease.

Conflicts of interest

The author is the developer of a deterministic structured reporting platform for musculoskeletal imaging. No external funding was received for this work, and no specific product performance claims are made.

This appendix provides a decision tree for question-driven imaging escalation in early seronegative peripheral inflammatory arthritis. The framework operationalizes the XR-first approach described in the main text, with explicit decision nodes for escalation to US-PD and selective MRI.

Stage descriptions

Order language templates

The following templates specify the clinical question and discriminative target for each modality.

Key principles

• XR-first. Always establish structural baseline before advanced imaging. XR provides distribution, symmetry, and mimic context that guides all subsequent interpretation.
• Question-driven. Escalate only when the imaging answer would change clinical management. Avoid reflexive escalation without a specific discriminative question.
• Targeted ordering. Specify the clinical question in order language. Generic requests yield generic reports; question-focused requests yield actionable answers.
• Selective MRI. Reserve MRI for BME/osteitis questions, deep anatomy, and persistent discordance. Routine MRI in all early arthritis is not evidence-supported (ARCTIC, IMAGINE-RA).
• Synthesis over accumulation. The goal is not maximum imaging but integrated phenotype probability and risk context that informs DMARD posture within treat-to-target care.

Quick reference — escalation triggers

Table A1.Quick reference — escalation triggers and expected answers.

This mini-atlas provides schematic illustrations for the five imaging patterns described in Section 6 of the main manuscript. Each pattern page includes anatomical localization, key morphological features, phenotype direction, principal mimics, and recommended next-test escalation.

How to use this atlas. When interpreting imaging in early seronegative peripheral inflammatory arthritis, match observed findings to the pattern templates below. Use the topography, morphology, and associated features to estimate phenotype probability, then follow the ‘best next test’ guidance for question-driven escalation.

Pattern A — marginal microerosion candidates at classic RA sites

Pattern B — enthesitis-adjacent cortical changes

Pattern C — periosteal reaction and fluffy new bone

Pattern D — DIP involvement with nail changes

Pattern E — asymmetric oligoarticular distribution

Quick reference summary

Table B1.Quick reference — five patterns for early seronegative peripheral inflammatory arthritis.

This checklist identifies required and recommended report elements for peripheral joint imaging in early seronegative inflammatory arthritis. The goal is to transform imaging reports from generic normal/abnormal statements into actionable clinical documents that support phenotype anchoring and question-driven escalation.

Target audience. Radiologists, musculoskeletal imaging specialists, rheumatologists ordering imaging, and quality improvement teams developing structured reporting templates.

1. Critical report elements — “do not omit” list

The following elements should be addressed in every peripheral imaging report for suspected early inflammatory arthritis. Omission reduces clinical utility and may contribute to diagnostic inertia.

Table C1.Critical report elements that should not be omitted in early seronegative IA imaging.

2. Plain radiography (XR) checklist

Indication. First-line imaging for suspected peripheral inflammatory arthritis; structural baseline; distribution mapping.

Table C2.Plain radiography (XR) reporting checklist.

3. Ultrasound with power Doppler (US-PD) checklist

Indication. Activity localization; tenosynovitis / enthesitis assessment; confirmation of inflammatory drivers at symptomatic sites.

Table C3.Ultrasound with power Doppler (US-PD) reporting checklist.

4. MRI checklist

Indication. BME/osteitis risk stratification; deep anatomy assessment; resolution of clinical-imaging discordance.

Table C4.MRI reporting checklist.

5. Confidence and uncertainty language

Reports should communicate diagnostic certainty appropriately. The following language gradations help calibrate clinical expectations.

Table C5.Confidence language gradations for imaging reports.

6. Common report pitfalls to avoid

Table C6.Common report pitfalls and preferred alternatives.

7. The actionable report

An actionable imaging report for early seronegative inflammatory arthritis should enable the clinician to:

• Update phenotype probability — is this RA-like, SpA-like, or mimic-favoring?
• Localize findings — which joints, which surfaces, what distribution?
• Assess activity context — is there objective inflammation (PD, BME) or just structural baseline?
• Screen for mimics — are there degenerative, crystal, or mechanical features?
• Plan next steps — is escalation to US-PD or MRI indicated, and what question should it answer?

Template-enforced, descriptor-complete reporting pipelines can systematically prompt these elements and improve consistency across readers and timepoints.

1. The reproducibility problem in radiographic reporting

Radiographic interpretation in inflammatory arthritis is subject to well-documented inter-reader and intra-reader variability. Studies of established scoring systems (Sharp, van der Heijde, Larsen) report inter-reader agreement coefficients that, while acceptable for group-level research, may translate into clinically meaningful discordance at the individual-patient level. In early seronegative disease, where findings are subtle and phenotype is uncertain, this variability is magnified.

Sources of variability include:

• Search behavior differences: which joints and surfaces are systematically evaluated.
• Threshold calibration: what constitutes a “positive” finding (e.g., erosion vs. cortical irregularity).
• Descriptor selection: how findings are named and characterized.
• Contextual weighting: how distribution, co-features, and clinical information influence interpretation.
• Report completeness: which elements are documented versus omitted.

For longitudinal monitoring — where detection of change is the clinical objective — variability compounds: a difference between timepoints may reflect true progression, regression, or simply inter-reader noise.

2. Structured reporting as a partial solution

Structured reporting — using templates, required fields, and standardized terminology — addresses some sources of variability by enforcing completeness and descriptor consistency. Organizations including the RSNA and ACR have promoted structured reporting initiatives across multiple imaging domains.

Benefits of structured reporting:

• Reduced omission bias (required fields ensure that key elements are addressed).
• Terminology standardization (controlled vocabularies reduce ambiguity).
• Improved data extractability (structured fields enable downstream analytics).
• Communication clarity (referring clinicians receive predictable, complete information).

Limitations of conventional structured reporting:

• Templates do not eliminate threshold variability (readers still differ in what they call “positive”).
• Completion does not ensure accuracy (a required field can be filled incorrectly).
• Contextual integration remains reader-dependent (pattern recognition is not enforced).
• Longitudinal comparability requires consistent application across timepoints.

3. Beyond templates — reproducibility layers

A more comprehensive approach to reproducibility extends beyond templates to incorporate what may be termed a “reproducibility layer” — a processing framework that enforces consistency at multiple stages of interpretation. Such systems aim to ensure that identical input data, processed through the same pipeline, yield identical output — a property often described as deterministic processing.

Conceptual components of a reproducibility layer may include:

• Input validation: verification that required data elements (projections, technical parameters, clinical context) are present and adequate before interpretation proceeds.
• Descriptor completeness checks: systematic confirmation that all required anatomical regions and finding categories have been addressed.
• Controlled vocabulary enforcement: restriction of descriptors to a defined lexicon, preventing synonymous but inconsistent terminology.
• Internal consistency verification: logic checks that flag contradictory findings or implausible combinations.
• Pattern-integration rules: structured logic for combining individual findings into pattern-level assessments.
• Longitudinal alignment: mechanisms to ensure that serial examinations are compared using consistent methodology.
• Audit-trail generation: documentation of processing steps to support reproducibility verification and regulatory review.

4. Architectural considerations

Platforms implementing reproducibility layers may adopt various architectural approaches. Key design considerations include the balance between deterministic and probabilistic components, the use of validation gating, separation of clinical and research outputs, and the depth of provenance/audit capability.

4.1 Deterministic vs. probabilistic processing

Probabilistic systems (including many AI/ML approaches) may yield different outputs for identical inputs due to model stochasticity, version drift, or training-data evolution. Deterministic systems, by contrast, are designed to produce identical outputs for identical inputs, supporting strict reproducibility requirements. Hybrid architectures may use probabilistic components for detection while applying deterministic rules for classification and reporting.

4.2 Validation gating

Sequential validation gates can ensure that each processing stage completes successfully before subsequent stages proceed. This architecture prevents incomplete or inconsistent data from propagating through the pipeline and generating unreliable outputs. Gate failures can trigger specific remediation pathways (e.g., request for additional projections, repeat exam, or manual review flags).

4.3 Separation of clinical and research outputs

Dual-layer architectures may separate clinical reporting (designed for immediate patient care) from research analytics (designed for standardized endpoints and longitudinal analysis). This separation allows clinical reports to retain appropriate hedging and contextualization, while research outputs maintain strict quantitative consistency and fixed terminology.

4.4 Provenance and audit capability

For regulatory and quality-assurance purposes, platforms may maintain comprehensive audit trails documenting input data, processing parameters, intermediate results, and final outputs. This provenance tracking supports reproducibility verification, error investigation, and regulatory submission requirements.

5. Implementation landscape

Several approaches to structured and reproducible radiographic reporting exist in the current landscape.

Table D1.Approaches to structured and reproducible radiographic reporting.

6. Illustrative architecture — deterministic validator-gated platform

To illustrate how the concepts described above may be implemented, this section describes design principles for a deterministic, validator-gated structured reporting platform applicable to musculoskeletal imaging in inflammatory arthritis. This description is conceptual and reflects one possible implementation pathway rather than a specific commercial product.

Core design principles

• Deterministic descriptor processing: identical input data processed through the platform yields identical outputs, supporting longitudinal comparability and audit requirements.
• Validator-governed architecture: sequential validation stages verify data integrity, projection adequacy, descriptor completeness, and internal consistency before outputs are generated.
• Dual-layer output separation: clinical reports are separated from research analytics, allowing appropriate context and wording for each use case.
• Controlled vocabulary: standardized terminology for anatomical localization, morphological description, and pattern classification.
• Audit-trail generation: provenance tracking designed to support regulatory review and reproducibility verification.

Potential applications

• Clinical-trial imaging endpoints, where reproducibility affects sample-size requirements and endpoint reliability.
• Longitudinal disease monitoring, where detection of true change requires minimizing measurement noise.
• Registry or observational cohorts, where consistent descriptors and export structures enable pooled analytics and phenotyping.
• Regulatory submissions, where deterministic processing and auditability may support approval pathways.

Limitations and considerations

Platform-based reporting does not eliminate the need for clinical judgment in ambiguous cases. Deterministic architecture is designed for scenarios where reproducibility is prioritized; applications requiring probabilistic confidence intervals or AI-driven pattern discovery may benefit from complementary approaches. Implementation requires workflow integration, user training, and clear governance over versioning and change-management.

7. Regulatory context

Emerging regulatory frameworks increasingly emphasize transparency and auditability in medical software and AI-based systems. Deterministic architectures with comprehensive audit trails may align with these evolving requirements. Relevant regulatory themes include:

• Software quality standards: international standards addressing software quality, reliability, and risk management in regulated clinical contexts.
• AI transparency requirements: regulatory expectations for explainability and auditability for AI/ML-based medical devices.
• Clinical-trial standards: imaging endpoints in clinical trials may require documented reproducibility, locked pipelines, and traceable changes.
• Reproducibility verification: regulators may request evidence that identical inputs produce identical outputs across processing instances, environments, and software versions.

Examples of relevant frameworks include international software quality standards, emerging AI regulation in the EU and US, and consensus efforts such as OMERACT imaging reproducibility initiatives. This appendix does not interpret or restate specific regulations, but highlights design features that may be relevant when aligning structured reporting platforms with regulatory expectations.

8. Future directions

The evolution of structured reporting platforms in inflammatory arthritis imaging may be shaped by several trends:

• Regulatory convergence: increased attention to AI transparency and reproducibility may favor deterministic or explainable architectures over opaque probabilistic systems.
• Cross-modality integration: platforms may expand to integrate radiography, ultrasound, and MRI within unified reproducibility frameworks.
• Clinical-trial standardization: pharmaceutical sponsors and CROs may increasingly require reproducible imaging endpoints, driving broader adoption of structured, validator-governed platforms.
• Longitudinal analytics: emphasis on detecting true change over time may prioritize architectures with strong longitudinal alignment and comparability contracts.
• Interoperability standards: development of common data models and exchange formats may enable platform outputs to integrate with electronic health records, registries, and research databases.

In early seronegative inflammatory arthritis, where subtle imaging changes carry disproportionate diagnostic and prognostic weight, such platforms may help convert fragile, reader-dependent impressions into reproducible, auditable descriptors that better support both clinical and research decision-making.

1. Rationale for separation

This manuscript has focused exclusively on peripheral joint and periarticular imaging patterns in early seronegative inflammatory arthritis. Axial skeleton imaging — including sacroiliac joints, spine, and pelvic/hip structures — is intentionally deferred to a companion paper for the following reasons.

1.1 Distinct lesion taxonomy

Axial inflammatory disease involves a fundamentally different lesion vocabulary: sacroiliitis (inflammatory, structural), syndesmophytes, corner inflammatory lesions (Romanus lesions), facet arthropathy, discovertebral changes, and entheseal ossification patterns (e.g., bridging syndesmophytes). These findings require different detection strategies, pattern recognition frameworks, and mimic considerations than peripheral erosions, synovitis, and enthesitis-adjacent changes.

1.2 Different modality priorities

While peripheral disease follows an XR-first pathway with US-PD as the primary activity modality, axial spondyloarthritis imaging prioritizes MRI for early detection of sacroiliac and spinal inflammation. The ASAS (Assessment of SpondyloArthritis international Society) classification criteria explicitly incorporate MRI sacroiliitis as an imaging arm. Radiographic sacroiliitis (modified New York criteria) remains relevant but detects later structural change. This modality inversion — MRI-first rather than XR-first — requires a distinct escalation framework.

1.3 Separate clinical pathways

Early axial spondyloarthritis often presents with inflammatory back pain, morning stiffness, and sacroiliac tenderness rather than peripheral joint swelling. While overlap exists (peripheral manifestations in axSpA, axial involvement in PsA), the diagnostic reasoning, referral patterns, and treatment considerations differ. Combining peripheral and axial frameworks in a single manuscript would dilute the operational focus and exceed practical length constraints.

1.4 Biomechanical considerations

Axial structures are subject to different biomechanical forces than peripheral joints. Degenerative mimics in the spine (facet arthrosis, disc degeneration, osteophytes) have different morphology and distribution than peripheral OA. The hip presents unique challenges as a transitional zone where both peripheral-type synovitis and axial-type enthesopathy/structural change may occur.

2. Planned companion paper scope

The companion manuscript, planned for submission approximately 6–12 weeks following acceptance of the current paper, will address:

Table E1.Planned scope of companion paper on axial and pelvic imaging patterns.

3. Bridging concepts

Several concepts from the current manuscript will carry forward to the axial companion paper:

• Imaging-first phenotype anchoring: the same probability-updating framework applies, using axial imaging patterns to strengthen or weaken axSpA phenotype probability in patients with inflammatory back pain.
• Question-driven escalation: escalation from XR to MRI (or from limited to comprehensive MRI) should be driven by specific clinical questions, not reflexive imaging.
• Explicit mimic management: axial mimics (degenerative SI change, DISH, mechanical back pain) require the same systematic checkpoint approach as peripheral mimics.
• Structured reporting: the checklist approach (required elements, descriptor options, confidence language) will be adapted for axial imaging contexts.
• Reproducibility emphasis: deterministic processing and validation gating concepts apply equally to axial imaging interpretation.

4. Cross-referencing guidance

For readers encountering patients with both peripheral and axial features:

• Peripheral-dominant presentation with back pain: apply the current manuscript’s framework for peripheral imaging; consider axial imaging if inflammatory back pain criteria (ASAS) are met and axial findings would change management.
• Axial-dominant presentation with peripheral symptoms: the companion paper will address primary axial assessment; peripheral imaging as described here may clarify peripheral manifestations.
• Mixed presentation (e.g., suspected PsA with axial involvement): both frameworks may be applicable; phenotype anchoring should integrate peripheral and axial findings.
• Hip involvement: the companion paper will address hip imaging in detail given its transitional anatomy; for now, consider the hip as potentially requiring both peripheral-type (synovitis) and axial-type (enthesopathy, structural) assessment.

5. Anticipated timeline

• Manuscript drafting: in progress.
• Internal review: 4–6 weeks following current manuscript acceptance.
• Target submission: 6–12 weeks following current manuscript acceptance.
• Target venue: same or related journal to facilitate cross-referencing.

The two manuscripts are designed to function as complementary components of a comprehensive imaging framework for early seronegative inflammatory arthritis, addressing peripheral and axial presentations with consistent methodology and shared conceptual foundations.