Literature Search Strategy: Autoimmune Dermatoses and Genodermatoses
Internal document. Part of Work-stream 3 (Pillar 1 literature) within task-3b5. Not shipped to BSI.
Status
COMPLETE (2026-04-20). All primary blocks (A–D), all supplementary searches (S1–S6), and full-text retrieval for the five immediate-priority Group A papers are complete. The audit-visible deliverable (evidence-package/pillar-1-literature.md) has been written, reviewed by BSI clinical auditor and Celine/Saray pillar-mapping agents, and revised. Next step: full-text retrieval for the remaining abstract-appraised load-bearing Group A papers (priority list in result.mdx §Full-Text Retrieval).
Purpose
This document records the search strategy executed to support Ingredient 1 of the triangulation package — the Pillar 1 Valid Clinical Association (VCA) literature review establishing that image-based clinical recognition of autoimmune dermatoses and genodermatoses is an accepted clinical standard with published diagnostic performance benchmarks.
The literature review targets the regulatory gap A.2.C5: both §6.5(e) declarations for autoimmune diseases (Gap 4) and genodermatoses (Gap 5) currently rest only on passive surveillance, which violates MDCG 2020-6 §6.4. Pillar 1 VCA literature is required to establish that image-based recognition of these sub-populations is clinically valid — before any discussion of the device's own performance.
Why a Multi-Block Search Structure
A single broad search (e.g., "autoimmune dermatoses + AI") would conflate heterogeneous topics and return unmanageable result volumes. The conditions covered span two fundamentally different disease families — autoimmune dermatoses (immunologically driven, often systemic) and genodermatoses (monogenic, structural, often rare) — with distinct imaging signatures, different diagnostic challenge profiles, and distinct AI/dermoscopy literatures.
The search was structured into four thematic blocks (A–D), each decomposed into sub-searches. This approach allows signal separation (each sub-search has a clear target), coverage control (gaps between sub-searches can be identified and filled), and appraisal tractability (focused result sets are easier to appraise than a 10,000-result dump).
Primary Block Structure
Block A: Autoimmune dermatoses — image-based and AI-assisted recognition
Target: peer-reviewed literature on dermoscopy, AI, or CNN-based recognition of the principal autoimmune dermatoses in the device's indication scope (dermatomyositis, lupus erythematosus, pemphigus group, bullous pemphigoid, lichen planus, morphea, vasculitis).
A1 (21 results): Inflammatory myopathies/dermatomyositis + machine learning. Captured AI/ML biomarker studies for dermatomyositis and erythematous-squamous skin disease classification algorithms.
A2 (18 results): Dermoscopy for clinically challenging inflammatory skin conditions. Captured dermoscopy studies for discoid lupus erythematosus (DLE), nail lichen planus, psoriasis, and inflammatory nail disorders across phototypes.
A3 (10 results): Deep learning / CNN architectures for multi-class skin disease recognition, including autoimmune conditions (bullous pemphigoid, lichen planus, psoriasis). Captured performance-benchmarked classification studies.
Block B: Genodermatoses — image recognition and diagnostic criteria
Target: literature on dermoscopy, AI, or clinical image-based recognition of the genodermatoses in scope (ichthyoses, epidermolysis bullosa, neurofibromatosis type 1 cutaneous manifestations, tuberous sclerosis cutaneous features, Gorlin syndrome/BCNS).
B1 (2 results): Deep learning for neurofibroma and NF-associated skin tumour diagnosis. Very specific search; low yield reflects limited literature on AI-specific NF1 cutaneous recognition.
B2 (9 results): Genodermatoses + dermoscopy / trichoscopy / clinical recognition. Captured ichthyosis phenotyping, Netherton syndrome trichoscopy, monilethrix dermoscopy, Gorlin syndrome, and café-au-lait macule characterisation in NF1.
B3 (79 results): Computational pathology and AI for skin tumours including basal cell carcinoma — relevant as the primary skin manifestation of Gorlin syndrome. Broader search capturing systematic reviews and landmark studies in AI-assisted skin cancer detection.
Block C: AI for inflammatory and rare dermatoses — broader sweep
Target: AI/ML studies broadly applied to inflammatory dermatoses and rare/genodermatoses, to supplement blocks A and B and capture papers that do not name a specific condition in the title.
C1 (12 results): AI/machine learning for inflammatory dermatoses broadly. Captured reviews, meta-analyses, and individual studies on psoriasis AI, indirect immunofluorescence (IIF) pattern recognition for autoimmune dermatoses, inflammatory skin disease severity assessment, and pyoderma gangrenosum classification.
C2 (3 results): Machine learning for rare skin diseases / genodermatoses specifically. Very low yield — reflects genuinely sparse literature but confirms that few-shot learning approaches are emerging for rare genodermatoses.
Block D: Clinical diagnosis standards and criteria for specific conditions
Target: clinical consensus documents, diagnostic criteria, and clinical recognition studies for specific autoimmune and genodermatoses conditions — to establish that image-based clinical recognition is an accepted standard-of-care approach.
D1 (85 results): Morphea / localised scleroderma clinical recognition. Broad search; first visible results were treatment-focused or neuroimaging; no dermoscopy-specific papers in visible results.
D2 (54,444 results): Vasculitis + dermatology diagnosis. Search too broad — parenthesis operator was ignored by PubMed, removing the intended scope restriction. First results page only reviewed; remainder unusable.
D3 (14,974 results): Failed search. PubMed ignored the opening parenthesis in the query, producing an unrelated broad search. All D3 results excluded.
D4 and D5: PDFs not delivered. Coverage from those sub-searches is unknown.
D6 (732 results): Neurofibromatosis type 1 diagnostic criteria. Captured the two key international consensus documents (I-NF-DC) updating NF1 and NF2/schwannomatosis diagnostic criteria, which anchor the clinical recognition standards for this genodermatosis.
Supplementary Searches S1–S6
Six supplementary searches were executed to address gaps identified after the primary block review. All six complete as of 2026-04-20.
S1: Pemphigus and Bullous Pemphigoid Clinical Guidelines
Method: The original PubMed query returned 337,821 results (too broad; abandoned). Five named guideline papers were retrieved by direct author/title search.
Yield: 5 papers appraised, all from full text.
- A-12 Hertl 2015 (EDF/EADV S2 pemphigus guideline) — 19/21 INCLUDE Group A
- A-13 Borradori 2022 (EADV S2K BP guideline) — 19/21 INCLUDE Group A
- A-15 Venning 2012 (BAD BP guideline) — 19/21 INCLUDE Group A
- A-14 Schmidt 2021 (EADV S3 MMP guideline) — 17/21 INCLUDE Group A
- A-16 Murrell 2008 (International Pemphigus consensus definitions) — 12/21 context only
Gap 5 (pemphigus/BP) status: CLOSED.
S2: Epidermolysis Bullosa
Method: PubMed, filters: Practice Guideline + Review, English, Humans. 86 results; first ~10 abstracts reviewed; authoritative GeneReviews chapters retrieved directly.
Yield: 3 Group A papers.
- A-17 Has et al. 2020 (Br J Dermatol, international EB consensus reclassification) — 19/21 INCLUDE Group A
- A-18 Pfendner & Lucky 2018 (GeneReviews NBK1125, JEB) — 16/21 INCLUDE Group A
- A-19 Lucky et al. 2025 (GeneReviews NBK1304, DEB) — 16/21 INCLUDE Group A
Gap 1 (EB) status: CLOSED.
S3: Tuberous Sclerosis Complex Cutaneous Features
Method: PubMed, filters: Free full text, English, Humans. 133 results; first ~4 abstracts reviewed.
Yield: 1 load-bearing Group A paper; 3 supporting context.
- A-20 Portocarrero 2018 (An Bras Dermatol, 2012 International Consensus TSC review) — 16/21 INCLUDE Group A
Gap 2 (TSC) status: CLOSED.
S4: Dermatomyositis Cutaneous Features
Method: PubMed, filters: Free full text + Books & Documents + Meta-Analysis + Practice Guideline + RCT + Review + SR, English, Humans. 118 results reviewed by title/abstract.
Yield: 4 Group A papers.
- A-21 Lundberg 2017 (Ann Rheum Dis, 2017 EULAR/ACR IIM classification criteria) — 20/21 INCLUDE Group A
- A-22 Ali 2025 (Semin Immunopathol, DM cutaneous features) — 17/21 INCLUDE Group A
- A-23 Didona 2023 (Ital J Dermatol Venerol, DM clinical review) — 15/21 INCLUDE Group A
- A-24 Chong & Werth 2022 (J Invest Dermatol, CDASI visual scoring) — 15/21 INCLUDE Group A
Gap 3 (DM) status: CLOSED.
S5: Morphea Dermoscopy
Method: PubMed, filters: Free full text + Books & Documents + Meta-Analysis + Practice Guideline + RCT + Review + SR, English, Humans. 39 results, all reviewed.
Outcome: No dermoscopy-specific morphea papers returned — publication-type filter excluded the small observational dermoscopy literature. Clinical features reviews found instead.
- A-27 Mertens 2017 (Am J Clin Dermatol, morphea and EF clinical update) — 15/21 INCLUDE Group A
Gap 4 (morphea) status: PARTIALLY CLOSED. Clinical recognition standard established; dermoscopy-specific evidence not captured. A search for "morphea" AND "dermoscopy" without publication-type filter is recommended for the SotA appendix in R-TF-015-011 (not required for the Pillar 1 VCA argument).
S6: Cutaneous Vasculitis
Method: PubMed, filters: same as S5. 104 results; first ~13 abstracts reviewed.
Yield: 2 Group A papers.
- A-25 Fraticelli 2021 (Intern Emerg Med, LCV diagnosis) — 15/21 INCLUDE Group A
- A-26 Chen 2024 (J Dermatol, cutaneous vasculitis patterns in AIDs) — 16/21 INCLUDE Group A
Gap 6 (cutaneous vasculitis) status: PARTIALLY CLOSED. Palpable purpura and cutaneous vasculitis skin patterns established as clinical recognition standard. Dermoscopy-specific evidence not captured (same filter issue as S5).
Boolean Search Strings
Boolean queries are documented below for all supplementary searches (S1–S6), recovered from the PubMed Advanced Search history (search-results/pubmed-search-history-2026-04-20.pdf). Primary block queries (A1–D6) were executed in a separate session (2026-04-19) for which the search history PDF is not available; their result PDFs are stored in search-results/primary/ and serve as the primary traceability record for those blocks.
Supplementary search queries
S1 — Pemphigus and Bullous Pemphigoid Clinical Guidelines (abandoned; replaced by direct named-paper retrieval)
("pemphigus" OR "bullous pemphigoid") AND ("guidelines" OR "consensus" OR "diagnostic criteria")
Returned 337,821 results — too broad; query abandoned. Five named guideline papers were retrieved directly by author/title search. No S1 result PDF.
S2 — Epidermolysis Bullosa
("epidermolysis bullosa") AND ("clinical features" OR "phenotype" OR "dermoscopy" OR "diagnostic criteria") AND ("skin" OR "cutaneous")
Filters: Practice Guideline, Review; English; Humans; date range 2011–2026. Returned 86 results.
S3 — Tuberous Sclerosis Complex Cutaneous Features
("tuberous sclerosis") AND ("skin" OR "cutaneous" OR "dermatolog*") AND ("diagnosis" OR "criteria" OR "phenotype" OR "imaging")
Filters: Free full text; Full text; English; Humans; date range 2014–2026. Returned 133 results.
S4 — Dermatomyositis Cutaneous Features
("dermatomyositis") AND ("skin" OR "cutaneous" OR "dermatolog*") AND ("diagnosis" OR "criteria" OR "dermoscopy" OR "clinical features")
Filters: Free full text; Full text; Books and Documents; Meta-Analysis; Practice Guideline; RCT; Review; Systematic Review; English; Humans; date range 2016–2026. Returned 118 results.
S5 — Morphea Dermoscopy
("morphea" OR "localized scleroderma") AND ("dermoscopy" OR "dermatoscopy" OR "clinical features" OR "diagnosis")
Filters: Free full text; Full text; Books and Documents; Meta-Analysis; Practice Guideline; RCT; Review; Systematic Review; English; Humans; date range 2006–2026. Returned 39 results.
S6 — Cutaneous Vasculitis
("cutaneous vasculitis" OR "leukocytoclastic vasculitis" OR "IgA vasculitis") AND ("diagnosis" OR "criteria" OR "dermoscopy" OR "clinical features")
Filters: Free full text; Full text; Books and Documents; Meta-Analysis; Practice Guideline; RCT; Review; Systematic Review; English; Humans; date range 2016–2026. Returned 104 results.
Full-Text Retrieval Summary
Five immediate-priority Group A papers were retrieved and appraised from full text (2026-04-20). All results recorded in result.mdx.
| Paper | Status | Key findings |
|---|---|---|
| Fathy et al. 2022 (A-06) | COMPLETE | Overall sens 88.6%, spec 71.2%, acc 83.5%. Follicular plugging: sens 95.8%, spec 100%, κ=0.868. Starburst pattern novel finding (39.3%). Score confirmed 17/21. |
| Tordjman et al. 2025 (A-01) | COMPLETE | SR PRISMA confirmed; 60 studies, 12,743 cases; phototype IV–VI onychoscopy criteria confirmed. Score confirmed 20/21. |
| Sun et al. 2021 (A-02) | COMPLETE — year corrected (was recorded as 2022) | Full citation: Sun Q et al. JAMA Dermatol. 2021;158(1):1–11. PMID 34851365. Standardised photographs used for genotype-phenotype mapping in 869/1000 kindreds. Score confirmed 20/21. |
| Legius et al. 2021 (A-03) | COMPLETE | I-NF-DC consensus confirmed; Delphi methodology confirmed; image-recognisable criteria (CALMs, freckling, cutaneous NF) confirmed. Score confirmed 20/21. |
| Bazzacco et al. 2024 (A-04) | COMPLETE — scope correction, score downgraded | Full text (DOI: 10.23736/S2784-8671.24.07825-3) confirmed scope is neoplastic-vs-inflammatory differential across 12 scenarios — NOT a standalone DLE/LP criteria review. DLE/LP appear only as inflammatory differentials from skin cancers. Score downgraded 18/21 → 14/21; paper demoted from load-bearing to supporting context in pillar-1-literature.md. |
Full-text retrieval complete (2026-04-20) except A-07. All load-bearing Group A papers have been retrieved and confirmed from full text. A-07 Jha et al. 2022 (Dermatology; IDS multicenter DLE cheilitis + oral LP dermoscopy, abstract score 17/21) was excluded from the evidence package — full text not publicly available. Oral LP has no retained load-bearing VCA anchor and is classified as Tier 3 (residual gap) in pillar-1-literature.md.
Updated Coverage Against Target Conditions
| Condition group | Target sub-conditions | Supplementary search | Final VCA status |
|---|---|---|---|
| Discoid lupus erythematosus | DLE | A2 (primary) | ✓ Closed — Fathy 2022 (sens/spec) |
| Dermatomyositis | DM cutaneous | S4 | ✓ Closed — Lundberg 2017 EULAR/ACR (20/21), 3 supporting anchors |
| Pemphigus group | Pemphigus, BP, MMP | S1 | ✓ Closed — 3 × 19/21 EADV/EDF/BAD guidelines + S3 MMP |
| Lichen planus (skin) | LP skin | A2 | ✓ Good — Sadhukhan 2025 (503 patients) as load-bearing; Bazzacco 2024 supporting context |
| Nail lichen planus | Nail LP | A2 | ✓ Excellent — Lim 2021 + Tordjman 2025 SR (12,743 cases) |
| Oral lichen planus | Mucosal LP | A2 | ✗ Gap — Jha 2022 excluded (full text unavailable); no load-bearing VCA anchor retained |
| Morphea | Localised scleroderma | S5 | ⚠ Partial — Mertens 2017 clinical features review (single abstract-appraised anchor at threshold) |
| Cutaneous vasculitis | LCV, cutaneous vasculitis | S6 | ⚠ Partial — palpable purpura + skin patterns established; dermoscopy-specific literature not captured |
| Ichthyoses | ARCI, lamellar, Netherton | B2 (primary) | ✓ Closed — Sun 2021 JAMA Dermatol (1,000 kindreds, full text) |
| Epidermolysis bullosa | EB simplex, JEB, DEB | S2 | ✓ Closed — Has 2020 international consensus + GeneReviews JEB/DEB |
| Neurofibromatosis type 1 | NF1 CALMs, cutaneous NF | D6 (primary) | ✓ Closed — Legius 2021 I-NF-DC (full text) |
| Neurofibromatosis type 2 | NF2 cutaneous features | D6 (primary) | ✓ Good — Plotkin 2022 I-NF-DC (abstract) |
| Tuberous sclerosis | TSC skin features | S3 | ✓ Closed — Portocarrero 2018 (2012 International Consensus review) |
| Gorlin syndrome (BCNS) | Multiple BCCs | B2, B3 | ✗ Gap — no disease-specific VCA anchor; BCC AI literature is Pillar 2 (SotA context only) |
| Netherton syndrome | Trichoscopy pattern | B2 | ⚠ Context only — single case report (Bittencourt 2015); no cohort or guideline |
| Monilethrix | Trichoscopy pattern | B2 | ⚠ Context only — single case report (Baltazard 2017) |
Search Execution Notes
- All primary searches (A1–A3, B1–B3, C1–C2, D1–D3, D6) executed on PubMed with filters: Free full text, Full text, English, Humans
- All supplementary searches (S1–S6) executed on PubMed; S1 replaced by direct named-paper retrieval after the automated query returned 337,821 results
- Primary block search result PDFs (A1–D6) should be placed in
search-results/primary/named{Block}{Sub-search}_literature search.pdfwhen provided. Note: D4 and D5 PDFs were never delivered; D3 results excluded (failed search); D2 reviewed at first-page only. The directorysearch-results/primary/has been created and is ready to receive these files. - Supplementary search result PDFs (S2–S6) are in
search-results/supplementary/namedS{N}_literature-search_{condition}.pdf. S1 was replaced by direct named-paper retrieval (no search-result PDF). - Individual manuscript PDFs for all Group A included papers are in
included-manuscripts/named{paper-id}_{first-author}_{year}.pdf - Database limitation: Embase and Cochrane Library were not searched; acknowledged in
pillar-1-literature.mdas a limitation; supplementary guideline-retrieval strategy partially compensates for autoimmune blistering disease anchors - Search execution dates: 2026-04-19 (primary blocks A–D); 2026-04-20 (supplementary searches S1–S6)
- D3 is a failed search (syntax error); its results are entirely excluded
- D2 is too broad (54,444 results); only the first visible page was reviewed
- D4 and D5 PDFs absent; coverage from those sub-searches unknown
- Where abstracts were sufficient for CRIT 1–7 appraisal, full texts were not accessed at this stage — see §Full-Text Retrieval Summary above for completed retrievals and remaining priority list in
result.mdx