Literature Review Results: Autoimmune Dermatoses and Genodermatoses (Pillar 1 VCA)

Internal document. Part of Work-stream 3 (Pillar 1 literature) within task-3b5. Not shipped to BSI. The appraisal outputs from this document feed into evidence-package/pillar-1-literature.md, which is the audit-visible deliverable.

---

Purpose

This document records the appraisal of all manuscripts retrieved by the 12 literature searches executed across blocks A–D (see search-strategy.md). Its purpose is to:

1. Record inclusion and exclusion decisions against CRIT 1–7 for every reviewed manuscript.
2. Identify which included papers are priorities for full-text retrieval.
3. Map included evidence against the target conditions listed in the task brief.
4. Assess remaining gaps requiring supplementary searches.

The appraisal serves Ingredient 1 of the triangulation strategy for A.2.C5: Pillar 1 Valid Clinical Association (VCA) literature establishing that image-based clinical recognition of autoimmune dermatoses and genodermatoses is an accepted clinical standard with published diagnostic performance benchmarks.

---

CRIT 1–7 Appraisal Framework

Criteria are scored 1–3 each; maximum total 21. Inclusion threshold: ≥12/21. Papers scoring ≥18/21 are priority papers; papers scoring 12–17/21 are standard includes; papers below 12/21 are excluded.

Criterion	Description	Score 1	Score 2	Score 3
CRIT1	Relevance to device indication	Irrelevant (wrong disease, modality, or task)	Partially relevant (relevant disease but indirect modality or tangential task)	Directly relevant (image-based recognition of a condition within scope, benchmarked performance)
CRIT2	Study design quality	Case report or expert opinion	Retrospective observational, cross-sectional, or small prospective cohort	Prospective multi-centre, RCT, systematic review, or meta-analysis
CRIT3	Reporting quality	Abstract only, no performance metrics	Performance metrics reported but without CIs or external validation	Full metrics with external validation or multi-centre testing against reference standard
CRIT4	Applicability to image-based clinical practice	Not image-based (genomics, blood-test, non-image ML)	Partially image-based (histology, limited dermoscopy scenarios, or non-standard clinical images)	Directly image-based (clinical photography, dermoscopy in routine use, or photographic diagnosis)
CRIT5	Evidence hierarchy	Case report, case series, or expert consensus (<5 studies)	Cohort, observational, or single-site RCT	Systematic review, meta-analysis, or landmark multi-site clinical study
CRIT6	Risk of bias	High (no reference standard, no blinding, retrospective with selection issues)	Moderate (retrospective single-site, small n, no external validation)	Low (external validation, prospective, blinded, large n)
CRIT7	Contribution to VCA claim	No contribution (therapeutic paper, wrong endpoint)	Indirect contribution (establishes disease burden or general AI performance)	Direct contribution (benchmarks image-based diagnostic accuracy for a condition in scope with sensitivity/specificity/AUC)

---

Block-by-Block Results

Block A: Autoimmune dermatoses — image-based and AI-assisted recognition

A1 (21 results): Inflammatory myopathies + ML

The search returned a heterogeneous mix. Several papers addressed inflammatory myopathy mechanisms and biomarker ML (gene expression, immunohistochemistry) rather than image-based skin recognition. The subset below addressed image-based or clinically applicable AI for autoimmune skin conditions.

#	First author	Year	Journal	Condition	Design	Total	Decision
A1-3	Cipriano	2025	An Bras Dermatol	Lichen planus, psoriasis, erythematous-squamous	Random forest, 366 patients, AUC 0.89–1.00	17	INCLUDE
A1-6	Vinayahalingam	2024	Clin Oral Investig	Oral lichen planus, OSCC, leukoplakia	Vision transformer, 4,161 images, OLP AUC 0.948	17	INCLUDE
A1-8	AlShareedah	2023	Sultan Qaboos Univ Med J	Systemic lupus erythematosus	CatBoost, 219 patients, AUC 0.95, cutaneous lupus feature	14	INCLUDE
A1-9	Eskandari	2024	Sci Rep	Psoriasis and lichen planus	ResNet-50, accuracy 89.07%, sensitivity 86.46%	18	INCLUDE (priority)
A1-10	Rewthamrongsris	2025	Int Dent J	Oral lichen planus	LLM vs CNN comparison, 1,142 images, CNN best	17	INCLUDE
A1-11	Hocke	2023	Front Immunol	Pemphigus, bullous pemphigoid, dermatitis herpetiformis	Computer-aided IIF classification, multi-substrate	19	INCLUDE (priority)
A1-1	Wischnewski	2025	Trends Pharmacol Sci	IIM mechanisms	Mechanisms review, ML mentioned tangentially	8	EXCLUDE
A1-2	Chen	2026	Adv Sci	Brain-imaging phenotypes, 18 AIDs	Mendelian randomisation, brain imaging — not skin	6	EXCLUDE
A1-4	Uehara	2021	Sci Rep	Parkinson's disease sebum RNA	Transcriptome ML — not skin recognition	5	EXCLUDE
A1-5	Pruthi	2024	Biomolecules	Oral squamous cell carcinoma	MicroRNA ML, mentions LP keyword only	8	EXCLUDE
A1-7	Wang	2024	Front Immunol	Dermatomyositis	ISG15 biomarker by IHC — gene expression not image	9	EXCLUDE

Note: Remaining 10 of 21 A1 papers not reviewed (beyond first 11 captured). The visible portion yielded 6 included papers. No additional review recommended for this block given adequate yield from reviewed subset.

A2 (18 results): Dermoscopy for challenging inflammatory conditions

This block was the most productive for the autoimmune VCA claim. Dermoscopy studies for DLE, nail lichen planus, and inflammatory cheilitis all directly establish image-based clinical recognition standards.

#	First author	Year	Journal	Condition	Design	Total	Decision
A2-1	Bazzacco	2024	Ital J Dermatol Venerol	DLE and LP as inflammatory differentials vs neoplastic lesions	Review of dermoscopy to differentiate neoplastic from inflammatory lesions across 12 scenarios (AK vs DLE, SCC vs hypertrophic LP, etc.); DLE/LP appear as inflammatory differentials from skin cancers — not a DLE/LP criteria review per se	14	INCLUDE (supporting context)
A2-3	Lim	2021	Acta Derm Venereol	Nail lichen planus, nail psoriasis, trachyonychia	Review, dermoscopic features of inflammatory nail disorders	18	INCLUDE (priority)
A2-4	Fathy	2022	Indian J Dermatol Venereol Leprol	Discoid lupus erythematosus	28 patients, dermoscopy + histopathology, sensitivity/specificity	17	INCLUDE (priority)
A2-5	Sadhukhan	2025	Indian J Dermatol Venereol Leprol	Lichen planus, lichen sclerosus, psoriasis	503 patients, prospective dermoscopy study	15	INCLUDE
A2-6	Tordjman	2025	Int J Dermatol	Nail lichen planus, nail psoriasis in Fitzpatrick IV–VI	SR, PRISMA, 60 studies, 12,743 cases	20	INCLUDE (priority)
A2-9	Jha	2022	Dermatology	DLE cheilitis, lichen planus of lips	IDS multicenter, dermoscopy criteria	17	INCLUDE
A2-2	De Luca	2020	G Ital Dermatol Venereol	Lentigo maligna	Primarily melanoma — not relevant to autoimmune scope	6	EXCLUDE
A2-7	Pereira	2024	Dermatology	Scalp lentiginous melanoma	Melanoma vs benign pigmented macules — not relevant	7	EXCLUDE
A2-8	Mulinari-Brenner	2017	An Bras Dermatol	Frontal fibrosing alopecia + LP pigmentosus	Case report, clinical challenge description	9	EXCLUDE (case report)

Note: Remaining A2 papers (10–18) not captured in extraction. Six included papers from first 9 reviewed is strong yield; block coverage for DLE, nail LP, and lips LP is good.

A3 (10 results): CNN/deep learning for multi-class skin disease

#	First author	Year	Journal	Condition	Design	Total	Decision
A3-1	Mathur	2021	Dermatol Ther	Bullous pemphigoid, psoriasis (20-class model)	CNN ensemble, 86.7% top-1 accuracy, Fitzpatrick-diverse	18	INCLUDE (priority)
A3-2	Gungor	2024	JAMA Ophthalmol	Giant cell arteritis / vasculitis (fundus)	DL multicenter 16 countries, AUC 0.97, external validation	18	INCLUDE (priority)
A3-3	Yu	2025	BMC Oral Health	Oral lichen planus (5-class OPMDs)	Dual-stage DL, AUC 0.9735, accuracy 0.9029	16	INCLUDE
A3-4	Vinayahalingam	2024	Clin Oral Investig	Oral lichen planus	Vision transformer, OLP AUC 0.948 (also in A1)	17	INCLUDE

Note: A3 appears to focus more on oral lichen planus and non-dermatological AI; remaining 6 papers not captured. Included papers nonetheless support VCA for lichen planus (oral and systemic) and bullous diseases.

---

Block B: Genodermatoses — image recognition and diagnostic criteria

B1 (2 results): Deep learning for neurofibroma and NF1

Both B1 results were captured (block was complete).

#	First author	Year	Journal	Condition	Design	Total	Decision
B1-1	Shi	2023	Am J Pathol	Neurofibroma, skin tumour pathology	Two-stage DL on whole slide images, NF + SK + BD	15	INCLUDE
B1-2	Matthies	2024	PLoS One	NF1 cutaneous neurofibroma	Raman spectroscopy + U-Net CNN, sensitivity 100%, specificity 97.3%	14	INCLUDE

Coverage note: Both NF1 papers use non-standard imaging modalities (histopathology slides, spectroscopy). Neither uses clinical photography or dermoscopy. They establish the feasibility of automated recognition of NF1 cutaneous lesions but do not yet anchor the VCA to routine clinical imaging. The D6 block (NF1 diagnostic criteria consensus) is the stronger VCA anchor for NF1.

B2 (9 results): Genodermatoses dermoscopy/trichoscopy/clinical recognition

#	First author	Year	Journal	Condition	Design	Total	Decision
B2-1	Sun	2021	JAMA Dermatol	Ichthyoses (ARCI, TGM1, KRT10, 32 genes)	Cohort 1000 kindreds, 10 years, standardised photographs	20	INCLUDE (priority)
B2-3	Bittencourt	2015	An Bras Dermatol	Netherton syndrome (ichthyosis linearis circumflexa)	Trichoscopy case report, bamboo hair pattern	12	INCLUDE (borderline)
B2-2	Baltazard	2017	Dermatol Online J	Monilethrix	Trichoscopy case report, necklace pattern	12	INCLUDE (borderline)
B2-4	Kosmidis	2023	Am J Case Rep	Gorlin syndrome (multiple BCCs)	Case report, diagnostic challenge	11	EXCLUDE (below threshold)

Coverage note: B2 yielded 3 included papers covering ichthyosis and two rare hair-shaft genodermatoses (Netherton, monilethrix). Gorlin syndrome is covered instead by B3 (BCC computational pathology). Epidermolysis bullosa and tuberous sclerosis remain unidentified in B2 — both are confirmed gaps requiring supplementary searches (see Section 6).

---

Block C: AI for inflammatory and rare dermatoses — broader sweep

C1 (12 results): AI/ML for inflammatory dermatoses broadly

All 12 results captured (block was complete).

#	First author	Year	Journal	Condition	Design	Total	Decision
C1-9	Hocke	2023	Front Immunol	Pemphigus, bullous pemphigoid, DH (IIF)	Computer-aided IIF classification (also A1-11)	19	INCLUDE (priority)
C1-10	Cai	2025	Br J Dermatol	Inflammatory skin disease severity (AI)	SR + meta-analysis, systematic	20	INCLUDE (priority)
C1-7	Silva	2025	An Bras Dermatol	Inflammatory dermatoses broadly	Review, AI landscape	18	INCLUDE (priority)
C1-2	Goessinger	2024	Am J Clin Dermatol	Psoriasis, image-based AI	Review, systematic	18	INCLUDE (priority)
C1-4	Busch	2025	BMJ Health Care Inform	Pyoderma gangrenosum, leg wounds	ML classification, wound imaging	17	INCLUDE
C1-6	Widiawaty	2026	F1000Res	Atopic dermatitis	Multimodal ML, image + clinical	15	INCLUDE
C1-11	Zaar	2020	Acta Derm Venereol	Skin disease diagnosis broadly	Online AI diagnostic accuracy	15	INCLUDE
C1-12	Doeleman	2025	J Invest Dermatol	Mycosis fungoides, inflammatory dermatoses	DL on H&E whole-slide images	15	INCLUDE
C1-1	Mulder	2021	Int J Mol Sci	Psoriatic arthritis vs psoriasis	Blood-based immune profiling ML — not image	8	EXCLUDE
C1-3	Zhang	2024	Skin Res Technol	Psoriasis key genes	RETRACTED	—	EXCLUDE
C1-5	Tang	2025	Front Immunol	Psoriasis housekeeping genes	Molecular bioinformatics — not image	6	EXCLUDE
C1-8	Zhou	2021	PLoS One	Psoriasis prediction	Lab tests, random forest — not image	7	EXCLUDE

C2 (3 results): ML for rare skin diseases / genodermatoses

All 3 results captured (block was complete).

#	First author	Year	Journal	Condition	Design	Total	Decision
C2-1	Schaefer	2020	Orphanet J Rare Dis	Rare diseases (incl. genodermatoses)	Scoping review, ML applications	18	INCLUDE (priority)
C2-2	Wen	2025	Math Biosci Eng	Rare skin diseases	Few-shot learning, adaptive calibration	17	INCLUDE (priority)
C2-3	Tschandl	2026	J Invest Dermatol	Skin cancer + simulators (benign inflammatory)	MILK10k dataset, multimodal hierarchical	18	INCLUDE

---

Block D: Clinical diagnosis standards and criteria

D1 (85 results): Morphea / localised scleroderma

The first visible results were predominantly treatment-focused (methotrexate resistance, mycophenolate mofetil) or addressed neuroimaging of Parry-Romberg/linear morphea. None of the first 4 visible papers addressed image-based clinical recognition with performance benchmarks.

#	First author	Year	Journal	Condition	Decision
D1-1	Pedowska	2022	Folia Morphol	Oral morphea, histopathological	EXCLUDE (histopathology, not skin imaging)
D1-2	Chalarca-Cañas	2024	An Bras Dermatol	Tattoo complications including morphea-like	EXCLUDE (tattoo complications)
D1-3	Gorolay	2025	Acad Radiol	Parry-Romberg / linear morphea, MRI	EXCLUDE (neuroimaging, not dermatology imaging)
D1-4	Martini	2021	Rheumatology	Juvenile localised scleroderma treatment	EXCLUDE (treatment study)

Assessment: D1's first visible results do not support the VCA for morphea. A dermoscopy-specific supplementary search for morphea is recommended (see Section 6).

D2 (54,444 results): Vasculitis + dermatology diagnosis

Search failed — parenthesis operator ignored by PubMed, producing an unmanageably broad vasculitis search. Only the first visible page was reviewed; results included vascular medicine papers without dermatology focus. No papers included from D2. Vasculitis VCA is partially covered by A3-2 (Gungor et al., giant cell arteritis/vasculitis optic imaging).

D3 (14,974 results): Failed search

Search failed — PubMed ignored the opening parenthesis, returning entirely irrelevant results (low back pain, frontotemporal dementia, intestinal tuberculosis). All D3 results excluded. A correctly formed supplementary search for the intended D3 query is recommended.

D4 and D5: PDFs absent

These sub-searches were not delivered. Coverage from D4 and D5 is unknown.

D6 (732 results): NF1 diagnostic criteria

The first two visible results were both I-NF-DC international consensus documents — the most authoritative sources for NF1 clinical recognition criteria.

#	First author	Year	Journal	Condition	Design	Total	Decision
D6-1	Legius	2021	Genet Med	NF1 + Legius syndrome diagnostic criteria	International consensus (I-NF-DC), Delphi, global experts + patients	20	INCLUDE (priority)
D6-2	Plotkin	2022	Genet Med	NF2 + schwannomatosis updated criteria	I-NF-DC international consensus	17	INCLUDE

Note: Legius et al. 2021 is the primary Group A VCA anchor for NF1: café-au-lait macules (CALMs), Lisch nodules, and cutaneous neurofibromas are image-recognisable diagnostic criteria revised by this consensus. Assigned ID A-03 in the consolidated list.

---

Supplementary Search S1: Pemphigus and Bullous Pemphigoid Clinical Guidelines

The original S1 PubMed query (("pemphigus" OR "bullous pemphigoid") AND ("guidelines" OR "consensus" OR "diagnostic criteria")) returned 337,821 results and was abandoned. Instead, five named guideline papers were retrieved directly by author/title search. All five are confirmed and appraised below.

ID	First author	Year	Journal	Document type	CRIT score	Decision
A-12	Hertl	2015	JEADV 29:405–414	Pemphigus S2 guideline, EDF/EADV, diagnosis + treatment	19/21	INCLUDE — Group A
A-13	Borradori	2022	JEADV 36:1689–1704	Bullous pemphigoid updated S2K guidelines, EADV	19/21	INCLUDE — Group A
A-15	Venning	2012	Br J Dermatol 167:1200–1214	BAD guidelines for management of BP, NHS Evidence accredited	19/21	INCLUDE — Group A
A-14	Schmidt	2021	JEADV 35:1926–1948	MMP S3 guidelines (diagnosis + management), EADV — Part II	17/21	INCLUDE — Group A
A-16	Murrell	2008	JAAD 58:1043–1046	International Pemphigus Committee consensus on definitions and endpoints	12/21	INCLUDE — Group A (context only)

CRIT4 note for all five: Clinical guidelines for autoimmune blistering diseases establish the visual clinical phenotype (flaccid blisters/erosions/Nikolsky for pemphigus; tense blisters on urticarial base for BP) as the basis for initial clinical suspicion — which is the device's functional role. Definitive confirmation requires DIF and serology; the device does not replace that confirmatory step. Score 2 applied to CRIT4 on this basis.

---

Supplementary Search S2: Epidermolysis Bullosa (86 results)

Filters applied: Practice Guideline, Review. Full list of 86 results; first ~10 abstracts reviewed.

Key findings: the search returned the 2020 international consensus reclassification of inherited EB (Has et al., Br J Dermatol) and two GeneReviews chapters (JEB and DEB). These establish the clinical phenotype of EB types as the basis for recognition: blistering with minimal trauma, healing with milia and scarring (DEB), granulation tissue and nail dystrophy (JEB). Definitive confirmation requires molecular genetic testing; clinical recognition from photographs is the standard first step.

ID	First author	Year	Journal	Condition	CRIT score	Decision
A-17	Has	2020	Br J Dermatol 183:614–627	Inherited EB — international consensus reclassification (292 citations)	19/21 (full text)	INCLUDE — Group A
A-18	Pfendner & Lucky	2018	GeneReviews [NBK1125]	Junctional EB — clinical features, diagnosis, management	16/21 (full text)	INCLUDE — Group A
A-19	Lucky et al.	2025	GeneReviews [NBK1304]	Dystrophic EB — clinical features, RDEB and DDEB	16/21 (full text)	INCLUDE — Group A
—	van Beek	2024	Front Immunol 15:1363032	AIBD state-of-the-art diagnosis (incl. EBA)	17/21	EXCLUDE — EBA/AIBD lab-based diagnosis paper; clinical features overlap with S1 scope (pemphigus/BP/MMP); primarily about DIF and serology, not image recognition
—	Hou	2021	Am J Clin Dermatol 22:801–817	EB investigational treatments	—	EXCLUDE — treatment focus
—	Marinkovich	2019	J Invest Dermatol 139:1221–1226	EB gene therapy	—	EXCLUDE — treatment focus

Gap 1 (EB) status: CLOSED. Three Group A VCA anchors establish that inherited EB (JEB, DEB, simplex EB) has clinically recognisable morphological features from photographs.

---

Supplementary Search S3: Tuberous Sclerosis Complex Cutaneous Features (133 results)

Filters applied: Free full text, English, Humans. First ~4 abstracts reviewed from 133 results.

Key findings: multiple review papers confirm that TSC cutaneous features — hypomelanotic macules (ash-leaf spots), facial angiofibromas, fibrous cephalic plaque, shagreen patches, ungual fibromas — are major diagnostic criteria per the 2012 International TSC Complex Consensus Conference. These are image-recognisable and are the first features to alert clinicians to the diagnosis. The Northrup/Krueger 2013 consensus paper (the primary reference for the 2012 revised criteria) likely appears deeper in the 133 results but is referenced by all the visible reviews.

ID	First author	Year	Journal	Condition	CRIT score	Decision
A-20	Portocarrero	2018	An Bras Dermatol 93:323–331	TSC review based on 2012 revised diagnostic criteria (59 citations)	16/21 (full text)	INCLUDE — Group A
—	Uysal & Şahin	2020	Turk J Med Sci 50:1665–1676	TSC review — 2012 criteria, mTOR, skin lesions	14/21 (abstract)	INCLUDE as context (below load-bearing threshold, backs A-20)
—	Wataya-Kaneda	2025	Keio J Med 74:42–51	TSC review — diagnosis, treatment, topical mTOR	14/21 (abstract)	INCLUDE as context
—	Togi	2022	Int J Mol Sci 23:11175	Genotype-phenotype of 283 Japanese TSC patients	15/21 (abstract)	INCLUDE as context — phenotype data supplements clinical recognition claim

Gap 2 (TSC) status: CLOSED. Portocarrero 2018 (A-20) describes the 2012 consensus criteria including cutaneous major criteria sufficient for VCA. Multiple supporting reviews confirm these criteria are in wide clinical use.

---

Supplementary Search S4: Dermatomyositis Cutaneous Features (118 results)

Filters applied: Free full text, Books and Documents, Meta-Analysis, Practice Guideline, RCT, Review, Systematic Review, English, Humans. Full 118 results reviewed by title/abstract.

Key finding: the 2017 EULAR/ACR classification criteria for idiopathic inflammatory myopathies (Lundberg et al., Ann Rheum Dis) is the primary VCA anchor — it assigns explicit probability weights to cutaneous features of DM (Gottron's papules as pathognomonic, heliotrope rash, V-sign, shawl sign, mechanic's hands) against a validated reference standard, with performance data from a multi-centre international cohort. This is the strongest Group A anchor in the entire S2–S6 set (20/21).

ID	First author	Year	Journal	Condition	CRIT score	Decision
A-21	Lundberg	2017	Ann Rheum Dis 76:1955–1964	2017 EULAR/ACR IIM classification criteria — cutaneous DM items with probability weights	20/21 (full text)	INCLUDE — Group A (primary anchor)
A-22	Ali	2025	Semin Immunopathol 47:32	DM focus on cutaneous features — Gottron, heliotrope, mechanic's hands, nail fold capillaries	17/21 (full text)	INCLUDE — Group A
A-23	Didona	2023	Ital J Dermatol Venerol 158:84–98	DM comprehensive review — cutaneous manifestations, serology, therapy	15/21 (full text)	INCLUDE — Group A
A-24	Chong & Werth	2022	J Invest Dermatol 142:936–943	CLE and DM — CDASI visual scoring tool	15/21 (full text)	INCLUDE — Group A
—	Davuluri	2024	Curr Opin Rheumatol 36:453	Calcinosis in DM	—	EXCLUDE — calcinosis only, not image recognition
—	Bellutti Enders	2017	Ann Rheum Dis 76:329	Juvenile DM consensus management	—	EXCLUDE — management focus
—	Paik	2025	Rheumatology 64:3288	IIM treatment guidelines	—	EXCLUDE — treatment focus
—	Wischnewski	2025	Trends Pharmacol Sci	IIM mechanisms	—	EXCLUDE — already in Block A1

Gap 3 (DM) status: CLOSED. Lundberg 2017 (A-21, 20/21) is the strongest single Group A paper in the supplementary searches; cutaneous DM features have formally validated probability weights in an international multi-centre study.

---

Supplementary Search S5: Morphea Dermoscopy (39 results)

Filters applied: Free full text, Books and Documents, Meta-Analysis, Practice Guideline, RCT, Review, Systematic Review, English, Humans. All 39 results reviewed.

Search outcome: no dermoscopy-specific papers found. The 39 results are dominated by systemic sclerosis treatment, IgA nephropathy, and management reviews. Dermoscopy of morphea is a small observational literature (case series, case-control studies) that does not appear as reviews or practice guidelines — these study types were excluded by the PubMed filter. The closest results are clinical reviews that describe the visual phenotype of morphea.

ID	First author	Year	Journal	Condition	CRIT score	Decision
A-27	Mertens	2017	Am J Clin Dermatol 18:491–512	Morphea and eosinophilic fasciitis — clinical update including skin features	15/21 (full text)	INCLUDE — Group A
—	Rongioletti	2018	G Ital Dermatol Venereol 153:208	Scleroderma — clinico-pathological correlation	13/21 (abstract)	INCLUDE as context (below load-bearing threshold)
—	All others	—	—	Systemic sclerosis, SSc nephrology, treatment reviews	<12	EXCLUDE

Gap 4 (morphea) status: PARTIALLY CLOSED.

The VCA for morphea rests on two observations: (a) morphea has characteristic image-recognisable clinical features — indurated ivory-white or violaceous plaques with a lilac ring at the active border, shiny surface, and skin tethering — that clinicians use for initial recognition before biopsy confirmation; (b) these features are described in standard clinical literature (Mertens 2017) and in the European EADV/EDF guidelines for localised scleroderma. The dermoscopy-specific literature (crystalline/white structures, fibrosis patterns, loss of follicular openings) exists in small observational studies not captured by the current filter but is available in the SotA literature. The VCA argument can proceed at clinical photography level; dermoscopy adds SotA context.

Residual note for CER author: a targeted search for "morphea" AND "dermoscopy" without publication-type filter would identify the Linthorst-Homan (2018) and related dermoscopy papers — this search is recommended before finalising the SotA section of R-TF-015-011 (but is not load-bearing for the Pillar 1 VCA argument itself).

---

Supplementary Search S6: Cutaneous Vasculitis (104 results)

Filters applied: Free full text, Books and Documents, Meta-Analysis, Practice Guideline, RCT, Review, Systematic Review, English, Humans. First ~13 abstracts reviewed.

Key findings: multiple papers establish that cutaneous vasculitis (LCV/IgA vasculitis/HSP) presents with palpable purpura as the pathognomonic clinical sign — an image-recognisable feature that is the established trigger for histopathological workup. The clinical recognition standard is well described; dermoscopy-specific papers were not returned (same filter issue as S5).

ID	First author	Year	Journal	Condition	CRIT score	Decision
A-25	Fraticelli	2021	Intern Emerg Med 16:831–841	LCV diagnosis and management — palpable purpura as leading clinical presentation	15/21 (full text)	INCLUDE — Group A
A-26	Chen	2024	J Dermatol 51:150–159	Cutaneous vasculitis in autoinflammatory diseases (FMF, DADA2, VEXAS) — skin pattern classification	16/21 (full text)	INCLUDE — Group A
—	Hetland	2017	Acta Derm Venereol 97:1160	HSP/IgA vasculitis literature review in a dermatology journal	13/21 (abstract)	INCLUDE as context — VCA for palpable purpura in IgA vasculitis
—	Parums	2024	Med Sci Monit 30:e943912	IgA vasculitis past/present/future	12/21 (abstract)	INCLUDE as context
—	Reamy	2020	Am Fam Physician 102:229	HSP rapid evidence review	12/21 (abstract)	INCLUDE as context
—	Vivarelli	2025	Pediatr Nephrol 40:533	IPNA IgA vasculitis nephritis guidelines	—	EXCLUDE — renal guideline, not skin recognition
—	Rajasekaran	2021	Am J Med Sci 361:176	IgA nephropathy — kidney disease	—	EXCLUDE — renal focus
—	González-Gay	2024	Best Pract Res Clin Rheumatol 38:101969	Genetics of vasculitis	—	EXCLUDE — genetics, not clinical recognition
—	Watts	2022	Nat Rev Rheumatol 18:22	Global epidemiology of vasculitis	—	EXCLUDE — epidemiology, not image recognition
—	Rothermel	2025	Am J Clin Dermatol 26:61	Urticarial vasculitis clinical algorithm	—	INCLUDE as context — UV skin presentation (urticarial wheals)
—	Murali	2021	Emerg Med Clin North Am 39:839	Abdominal pain mimics	—	EXCLUDE — not relevant

Gap 6 (cutaneous vasculitis) status: PARTIALLY CLOSED.

The VCA argument for cutaneous vasculitis rests on the same structural principle as pemphigus/BP: clinical visual features (palpable purpura, petechiae, livedo racemosa, urticarial wheals) trigger clinical suspicion and are described as pathognomonic or characteristic in multiple papers, even though definitive diagnosis requires histopathology. Fraticelli 2021 explicitly states "the leading clinical presentation of LCV is palpable purpura" — palpable purpura is an image-recognisable skin finding. Chen 2024 provides a J Dermatol-level classification of cutaneous vasculitis skin patterns in AIDs. The device's role is clinical recognition at this initial-suspicion level, not histopathological confirmation.

---

Block B3 (79 results): Computational pathology for skin tumours including BCC

B3 was too large to review in full (79 results). The first visible results were landmark papers in computational skin cancer AI. Two papers are extracted as context references for the Gorlin syndrome VCA (where BCC is the primary cutaneous manifestation):

#	First author	Year	Journal	Condition	Notes
B3-cx1	Campanella	2019	Nat Med	BCC and other cancers, whole slide images	Clinical-grade DL, AUC >0.98, 44,732 slides — establishes technical feasibility
B3-cx2	Jones (Cochrane)	2022	Lancet Digit Health	Skin cancer AI in primary care	SR 272 studies, melanoma 89.5% / BCC 87.6% mean accuracy

These context references support the technical feasibility claim for the Gorlin syndrome VCA but do not directly establish image-based clinical recognition criteria for Gorlin syndrome itself.

---

Consolidated Included Manuscript List

Important classification note (applied after Celine/Saray and BSI clinical auditor review).

The CRIT 1–7 matrix scores papers on study design quality, reporting quality, and applicability to image-based diagnosis. However, for the Pillar 1 VCA specifically, the relevant question under MDCG 2020-1 §4.2 is narrower: is the scientific association between image-based recognition of the condition and clinical management accepted by the broad medical community? That question is answered by clinical standards-of-care sources — international diagnostic consensus documents, dermoscopy criteria reviews, large phenotyping cohorts, professional-society guidelines. It is not answered by AI-performance benchmarking studies, which establish Pillar 2 (analytical performance) or Pillar 3 (reader-study) evidence instead.

The CRIT matrix as designed scores many AI-performance studies highly (because they have good study design, report metrics, and are applicable to image-based diagnosis). Those papers are valuable SotA evidence and belong in R-TF-015-011; they do not carry the Pillar 1 VCA argument. Accordingly, included papers are classified into two groups: Group A — Pillar 1 VCA anchors and Group B — SotA / technological-feasibility context (Pillar 2/3 support). Only Group A papers count towards the VCA sufficiency determination.

Group A: Pillar 1 VCA anchors (clinical-standard literature)

These are the papers that answer "is image-based clinical recognition of this condition an accepted clinical practice with known diagnostic criteria?" They are consensus documents, dermoscopy-criteria reviews, systematic reviews of clinical standards, and large phenotyping cohorts. They do not need AI performance numbers to qualify as VCA evidence.

Appraisal note: papers reviewed from abstract only are marked (abstract); CRIT3 and CRIT6 scores should be treated as provisional pending full-text confirmation.

ID	First author	Year	Journal	Condition	CRIT score	VCA anchor
A-01	Tordjman	2025	Int J Dermatol	Nail LP, nail psoriasis, Fitzpatrick IV–VI (SR, 60 studies, 12,743 cases)	20/21 (full text)	Nail lichen planus dermoscopy criteria — SoC across phototypes
A-02	Sun	2021	JAMA Dermatol	Ichthyoses (1000 kindreds, standardised photographs, 32 genes; Sun Q, Burgren NM, Cheraghlou S et al., JAMA Dermatol. 2021;158(1):1–11, PMID 34851365)	20/21 (full text)	Clinical phenotyping standard for ichthyosis — standardised clinical photographs used to map genotype-phenotype associations across 32 genes in 869/1000 kindreds
A-03	Legius	2021	Genet Med	NF1 + Legius syndrome (I-NF-DC international consensus)	20/21 (full text)	NF1 diagnostic criteria — CALMs, cutaneous NF as image-recognisable features
A-04	Bazzacco	2024	Ital J Dermatol Venerol	DLE and LP as inflammatory differentials vs neoplastic lesions (dermoscopy review; 12 scenarios: AK vs DLE, SCC vs hypertrophic LP)	14/21 (full text) — supporting context	Dermoscopy differentiates DLE from AK and LP from SCC — CRIT7 contribution is differential-discriminability, not standalone recognition-standard; demoted from load-bearing to supporting context after reviewer-agent audit
A-05	Lim	2021	Acta Derm Venereol	Nail LP, nail psoriasis, trachyonychia onychoscopy (review)	18/21 (full text)	Dermoscopy criteria for nail LP — established diagnostic practice
A-06	Fathy	2022	Indian J Dermatol Venereol Leprol	DLE dermoscopy, 28 patients, histopathological correlation — overall sens 88.6%, spec 71.2%, acc 83.5%; follicular plugging: sens 95.8%, spec 100%, acc 96.4%, κ=0.868; starburst pattern novel finding (39.3% prevalence)	17/21 (full text)	DLE dermoscopic criteria with validated sensitivity/specificity; follicular plugging is the highest-performing single feature
A-07	Jha	2022	Dermatology	DLE cheilitis + LP lips, IDS multicenter dermoscopy	17/21 (abstract — full text not publicly available; EXCLUDED from evidence package)	Excluded — full text not publicly accessible; abstract-appraised score unverifiable; removed from load-bearing list and from pillar-1-literature.md
A-08	Plotkin	2022	Genet Med	NF2 + schwannomatosis (I-NF-DC international consensus)	17/21 (full text)	NF2 updated diagnostic criteria — clinical recognition standards
A-09	Sadhukhan	2025	Indian J Dermatol	LP, lichen sclerosus, psoriasis dermoscopy (503 patients)	15/21 (full text)	LP dermoscopic criteria in prospective clinical series
A-10	Bittencourt	2015	An Bras Dermatol	Netherton syndrome, trichoscopy case report	12/21 (context only)	Trichoscopy pathognomonic for Netherton — single case, context only; not VCA load-bearing
A-11	Baltazard	2017	Dermatol Online J	Monilethrix trichoscopy case report	12/21 (context only)	Trichoscopy for monilethrix — single case, context only; not VCA load-bearing
A-12	Hertl	2015	JEADV 29:405–414	Pemphigus S2 guideline, EDF/EADV (full text)	19/21	Pemphigus clinical visual features (flaccid blisters, erosions, Nikolsky) as standard of care for recognition
A-13	Borradori	2022	JEADV 36:1689–1704	Bullous pemphigoid S2K guideline, EADV (full text)	19/21	BP clinical visual phenotype (tense blisters on urticarial base) as recognition standard
A-14	Schmidt	2021	JEADV 35:1926–1948	MMP S3 guideline, EADV — diagnosis + management (full text)	17/21	MMP cutaneous/mucosal features as diagnostic basis; scarring mucosal disease
A-15	Venning	2012	Br J Dermatol 167:1200–1214	BAD guidelines for BP, NHS Evidence accredited (full text)	19/21	BP clinical recognition criteria — national guideline level
A-16	Murrell	2008	JAAD 58:1043–1046	International Pemphigus Committee consensus on definitions (full text)	12/21 (context only)	Pemphigus lesion definitions; informs image-label taxonomy but not a recognition-standard anchor
A-17	Has	2020	Br J Dermatol 183:614–627	Inherited EB international consensus reclassification — full clinical+genetic taxonomy	19/21 (full text)	EB blistering morphology, subtypes, and clinical recognition criteria — primary EB VCA anchor
A-18	Pfendner & Lucky	2018	GeneReviews [NBK1125]	Junctional EB — clinical features, molecular diagnosis	16/21 (full text)	JEB clinical phenotype (blistering, granulation tissue, nail dystrophy) as basis for recognition
A-19	Lucky et al.	2025	GeneReviews [NBK1304]	Dystrophic EB — RDEB and DDEB clinical features	16/21 (full text)	DEB clinical phenotype (scarring blisters, pseudosyndactyly) as basis for recognition
A-20	Portocarrero	2018	An Bras Dermatol 93:323–331	TSC review based on 2012 revised International Consensus diagnostic criteria	16/21 (full text)	TSC cutaneous major criteria (angiofibromas, ash-leaf spots, shagreen patch) as image-recognisable diagnostic markers
A-21	Lundberg	2017	Ann Rheum Dis 76:1955–1964	2017 EULAR/ACR IIM classification criteria — DM cutaneous items with probability weights, multicenter	20/21 (full text)	Gottron's papules, heliotrope rash, V-sign as formally validated image-based DM diagnostic criteria
A-22	Ali	2025	Semin Immunopathol 47:32	DM cutaneous features — Gottron, heliotrope, mechanic's hands, nail fold changes	17/21 (full text)	Cutaneous DM features as primary clinical recognition standard
A-23	Didona	2023	Ital J Dermatol Venerol 158:84–98	DM comprehensive review — cutaneous manifestations, serology	15/21 (full text)	DM clinical manifestations review with dermatological perspective
A-24	Chong & Werth	2022	J Invest Dermatol 142:936–943	CLE and DM cutaneous assessment tools (CDASI visual scoring)	15/21 (full text)	CDASI visual scoring confirms standardised cutaneous DM feature assessment
A-25	Fraticelli	2021	Intern Emerg Med 16:831–841	LCV diagnosis and management — palpable purpura as pathognomonic presentation	15/21 (full text)	Palpable purpura as image-recognisable clinical trigger for vasculitis workup
A-26	Chen	2024	J Dermatol 51:150–159	Cutaneous vasculitis in autoinflammatory diseases — skin pattern classification	16/21 (full text)	Cutaneous vasculitis skin patterns (purpura, livedo racemosa) as clinical recognition criteria
A-27	Mertens	2017	Am J Clin Dermatol 18:491–512	Morphea and eosinophilic fasciitis — clinical update including skin recognition features	15/21 (full text)	Morphea indurated plaques with lilac ring as image-recognisable clinical standard

Load-bearing VCA anchors (A-01 through A-09 excluding A-04 and A-07, A-12 through A-15, A-17 through A-27 excluding context-only): 22 papers. A-04 (Bazzacco 2024) demoted from 15/21 load-bearing to supporting context (14/21) after reviewer-agent audit: the paper establishes dermoscopy differentiates DLE/LP from neoplastic mimics, which is a differential-discriminability claim, not a standalone recognition-standard claim; the CRIT7 contribution is adjacent to but not identical to the VCA claim. A-10 and A-11 are context only — case reports cannot establish that a clinical recognition standard is accepted by the broad medical community. The Netherton VCA requires either a cohort study or a professional-guideline citation from the ichthyosis literature.

Group B: SotA / technological-feasibility context (Pillar 2 and Pillar 3 support)

These papers establish that AI-based classification of the relevant conditions has been implemented and benchmarked, supporting the Pillar 2 analytical-performance claim and the broader SotA narrative for R-TF-015-011. They do not constitute Pillar 1 VCA evidence and must not be presented as such in the CER or SotA appendix.

ID	First author	Year	Journal	Condition	CRIT score	SotA contribution
B-01	Hocke	2023	Front Immunol	AIBD (pemphigus, BP, DH) — computer-aided IIF	19/21 (abstract)	AI classification of autoimmune blistering diseases; Pillar 2 SotA
B-02	Cai	2025	Br J Dermatol	Inflammatory skin severity — SR + meta-analysis	20/21 (abstract)	AI severity assessment feasibility; Pillar 2 SotA
B-03	Eskandari	2024	Sci Rep	DL for psoriasis and lichen planus, ResNet-50	18/21 (abstract)	Benchmarked AI accuracy for LP; Pillar 2 SotA
B-04	Mathur	2021	Dermatol Ther	CNN 20-class skin, bullous pemphigoid included	18/21 (abstract)	Multi-class CNN includes BP; Pillar 2 SotA
B-05	Gungor	2024	JAMA Ophthalmol	Giant cell arteritis vasculitis — fundus imaging	18/21 (abstract)	AI for autoimmune vasculitis — fundus domain, not skin; Pillar 2 analogy only
B-06	Goessinger	2024	Am J Clin Dermatol	Psoriasis image AI — review	18/21 (abstract)	Image AI for psoriasis; Pillar 2 SotA review
B-07	Silva	2025	An Bras Dermatol	AI for inflammatory dermatoses — review	18/21 (abstract)	Broad inflammatory AI landscape; Pillar 2 SotA review
B-08	Schaefer	2020	Orphanet J Rare Dis	ML for rare diseases — scoping review	18/21 (abstract)	ML for genodermatoses — feasibility; Pillar 2 SotA review
B-09	Tschandl	2026	J Invest Dermatol	MILK10k toolkit, skin cancer + simulators	18/21 (abstract)	Large-scale skin image AI; SotA context only
B-10	Cipriano	2025	An Bras Dermatol	Erythematous-squamous incl. LP, RF AUC 0.89–1.00	17/21 (abstract)	AI for LP; Pillar 2 SotA (LP-specific per-class performance needed from full text)
B-11	Vinayahalingam	2024	Clin Oral Investig	OLP vision transformer, AUC 0.948	17/21 (abstract)	AI for oral LP; Pillar 2 SotA
B-12	Rewthamrongsris	2025	Int Dent J	LLMs vs CNNs for OLP (1,142 images)	17/21 (abstract)	Model comparison for oral LP; Pillar 2/3 SotA
B-13	Busch	2025	BMJ Health Care Inform	ML for pyoderma gangrenosum, wound imaging	17/21 (abstract)	AI classification of PG; Pillar 2 SotA
B-14	Wen	2025	Math Biosci Eng	Few-shot learning for rare skin diseases	17/21 (abstract)	FSL feasibility for genodermatoses; Pillar 2 SotA
B-15	AlShareedah	2023	Sultan Qaboos Univ Med J	ML for SLE prediction, cutaneous features	14/21 (abstract)	Clinical ML for SLE (not image-based); SotA context
B-16	Yu	2025	BMC Oral Health	Dual-stage DL for OPMDs incl. OLP	16/21 (abstract)	AI for oral LP; Pillar 2 SotA
B-17	Widiawaty	2026	F1000Res	Multimodal ML for atopic dermatitis	15/21 (abstract)	ML for AD; broad inflammatory SotA
B-18	Zaar	2020	Acta Derm Venereol	Online AI diagnostic accuracy skin disease	15/21 (abstract)	Clinical AI accuracy benchmark; SotA
B-19	Doeleman	2025	J Invest Dermatol	DL for mycosis fungoides vs benign inflammatory, H&E WSI	15/21 (abstract)	AI for inflammatory DL (histopathology modality — not device input)
B-20	Shi	2023	Am J Pathol	DL for neurofibroma, whole-slide images	15/21 (abstract)	AI for NF1 (histopathology modality — not device input)
B-21	Matthies	2024	PLoS One	Raman spectroscopy + U-Net for NF1 neurofibroma	14/21 (abstract)	AI for NF1 (Raman modality — not device input)

Total: 22 Group A load-bearing VCA anchors (score ≥15/21, full text retrieved) + 5 Group A non-load-bearing (A-04 supporting context 14/21; A-07 excluded full text unavailable; A-10, A-11, A-16 context only 12/21) + 21 Group B SotA context papers. Two additional context references from B3 (Campanella 2019, Jones/Cochrane 2022) are noted separately.

---

Full-Text Retrieval Priority List

The following papers require full-text access to extract diagnostic performance benchmarks (sensitivity, specificity, AUC, accuracy), sample sizes, and reference standard used. Order reflects VCA priority.

Immediate priority (Group A VCA anchors — full text to confirm provisional CRIT3/CRIT6 scores)

COMPLETE (2026-04-20). All five immediate-priority papers retrieved and appraised from full text. Scores updated in consolidated Group A table above.

1. A-06 Fathy et al. 2022 (Indian J Dermatol) — COMPLETE. Full text confirmed. Overall sensitivity 88.6%, specificity 71.2%, accuracy 83.5%. Follicular plugging: sensitivity 95.8%, specificity 100%, accuracy 96.4%, κ=0.868. Starburst pattern: novel finding reported for the first time. CRIT3 and CRIT6 scores confirmed. Score remains 17/21.
2. A-01 Tordjman et al. 2025 (Int J Dermatol) — COMPLETE. Full text confirmed. SR with PRISMA compliance; 60 studies, 12,743 cases; onychoscopy criteria covered across phototypes IV–VI. Score confirmed 20/21.
3. A-02 Sun et al. 2022 (JAMA Dermatol) — COMPLETE (year corrected: 2021, not 2022). Full citation: Sun Q, Burgren NM, Cheraghlou S, Paller AS, Larralde M, Bercovitch L, Levinsohn J, Ren I, Hu RH, Zhou J, Zaki T, Fan R, Tian C, Saraceni C, Nelson-Williams CJ, Loring E, Craiglow BG, Milstone LM, Lifton RP, Boyden LM, Choate KA. JAMA Dermatol. 2021 Dec 1;158(1):1–11. PMID 34851365. Standardised clinical photographs used throughout for genotype-phenotype mapping (869/1000 kindreds had variants; 266 novel variants in 32 genes). Score confirmed 20/21.
4. A-03 Legius et al. 2021 (Genet Med) — COMPLETE. Full text confirmed. I-NF-DC consensus; Delphi methodology confirmed; CALMs, axillary/inguinal freckling, cutaneous NF as image-recognisable major criteria confirmed. Score confirmed 20/21.
5. A-04 Bazzacco et al. 2024 (Ital J Dermatol Venerol) — COMPLETE (scope correction applied). Full text (DOI: 10.23736/S2784-8671.24.07825-3) confirmed actual scope: differentiating neoplastic from inflammatory skin lesions across 12 scenarios. DLE appears only as an inflammatory differential from AK; LP appears only as an inflammatory differential from SCC. This is NOT a standalone DLE/LP dermoscopy criteria review. Score downgraded 18/21 → 14/21; paper demoted from load-bearing to supporting context in pillar-1-literature.md. Paper remains a valid Group A INCLUDE as it confirms dermoscopy differentiates DLE/LP from similar-appearing cancers.

Secondary priority (Group A VCA anchors — full text retrieval)

COMPLETE (2026-04-20) except A-07 (excluded). All secondary-priority Group A papers retrieved from full text. Scores remain unchanged from abstract appraisal.

6. A-05 Lim et al. 2021 (Acta Derm Venereol) — COMPLETE. Full text retrieved. Score 18/21 confirmed.
7. A-07 Jha et al. 2022 (Dermatology) — EXCLUDED. Full text not publicly available. Paper removed from Group A load-bearing list and from pillar-1-literature.md. Oral LP drops to Tier 3 (no load-bearing anchor retained).
8. A-09 Sadhukhan et al. 2025 (Indian J Dermatol Venereol Leprol) — COMPLETE. Full text retrieved. Score 15/21 confirmed.
9. A-21 Lundberg et al. 2017 (Ann Rheum Dis) — COMPLETE. Full text retrieved. Score 20/21 confirmed.
10. A-23 Didona et al. 2023 and A-24 Chong & Werth 2022 — COMPLETE. Full texts retrieved. Scores 15/21 confirmed.
11. A-08 Plotkin et al. 2022, A-20 Portocarrero et al. 2018, A-22 Ali et al. 2025, A-25 Fraticelli et al. 2021, A-26 Chen et al. 2024, A-27 Mertens et al. 2017 — COMPLETE. All full texts retrieved. Scores confirmed.

Secondary priority (Group B SotA papers — for SotA benchmark extraction in R-TF-015-011)

12. B-01 Hocke et al. 2023 (Front Immunol) — per-class performance for pemphigus, BP, DH
13. B-03 Eskandari & Sharbatdar 2024 (Sci Rep) — per-class accuracy LP vs psoriasis; confusion matrix
14. B-04 Mathur et al. 2021 (Dermatol Ther) — per-class sensitivity/specificity for BP
15. B-02 Cai et al. 2025 (Br J Dermatol) — meta-analysis subgroup data for specific inflammatory conditions
16. B-13 Busch et al. 2025 (BMJ Health Care Inform) — pyoderma gangrenosum classifier performance

Lower priority (abstract sufficient; Group B SotA narrative)

17. B-06 Goessinger et al. 2024 — psoriasis AI review
18. B-07 Silva et al. 2025 — inflammatory dermatoses AI review
19. B-08 Schaefer et al. 2020 — rare disease ML scoping review
20. B-11 Vinayahalingam et al. 2024 — OLP vision transformer
21. B-16 Yu et al. 2025 — OPMDs dual-stage classification
22. B-15 AlShareedah et al. 2023 — SLE prediction (abstract sufficient)

---

Evidence Gap Coverage Assessment

Autoimmune dermatoses

Condition	Target VCA requirement	Papers identified	Coverage
Discoid lupus erythematosus (DLE)	Dermoscopy criteria, diagnostic accuracy	Fathy 2022 (A-06)	✓ Good
Systemic lupus erythematosus	Cutaneous image recognition criteria	AlShareedah 2023 recategorised to Group B (clinical ML, not image-based)	✗ Gap — no Group A VCA anchor; DLE covered separately
Dermatomyositis	Cutaneous feature recognition	Lundberg 2017 (A-21) EULAR/ACR criteria; Ali 2025 (A-22); Didona 2023 (A-23); Chong & Werth 2022 (A-24)	✓ Excellent
Pemphigus group	Clinical consensus criteria	Hertl 2015 (A-12) EDF/EADV S2; Murrell 2008 (A-16, context only)	✓ Good
Bullous pemphigoid	Clinical consensus criteria	Borradori 2022 (A-13) EADV S2K; Venning 2012 (A-15) BAD	✓ Good
Mucous membrane pemphigoid (MMP)	Clinical recognition criteria	Schmidt 2021 (A-14) EADV S3	✓ Good
Lichen planus (skin)	Dermoscopy criteria	Sadhukhan 2025 (A-09); Bazzacco 2024 (A-04, supporting context only)	✓ Good (dermoscopy criteria established; single load-bearing anchor — Sadhukhan 2025 — supplemented by Bazzacco 2024 differential-mimics context)
Nail lichen planus	Onychoscopy criteria	Lim 2021 (A-05); Tordjman 2025 SR (A-01)	✓ Excellent
Oral lichen planus	Mucosal dermoscopy criteria	None — Jha 2022 (A-07) excluded (full text unavailable)	✗ Gap — no load-bearing VCA anchor retained
Morphea / scleroderma	Image recognition of cutaneous features	Mertens 2017 (A-27) — clinical features review	⚠ Partial — clinical recognition standard established; single abstract-appraised anchor at 15/21 threshold; dermoscopy-specific literature not captured (observational studies excluded by filter)
Cutaneous vasculitis	Skin-image recognition criteria	Fraticelli 2021 (A-25); Chen 2024 (A-26)	⚠ Partial — palpable purpura/skin patterns established as clinical recognition standard; dermoscopy not captured
Pyoderma gangrenosum	Wound image classification AI	Busch 2025 (Group B)	⚠ SotA context only — no Group A VCA anchor

Genodermatoses

Condition	Target VCA requirement	Papers identified	Coverage
Ichthyoses (ARCI, lamellar)	Clinical image recognition, genotype-phenotype	Sun 2021 (A-02) — 1000 kindreds, standardised photos	✓ Good
Netherton syndrome	Trichoscopy pathognomonic feature	Bittencourt 2015 (A-10, context only)	⚠ Context only — single case report, not a load-bearing VCA anchor
Monilethrix	Trichoscopy recognition	Baltazard 2017 (A-11, context only)	⚠ Context only — single case report
Epidermolysis bullosa	Clinical image recognition	Has 2020 (A-17) international consensus; GeneReviews JEB (A-18) and DEB (A-19)	✓ Good
Neurofibromatosis type 1 (CALMs, cutaneous NF)	Clinical recognition criteria, AI	Legius 2021 (D6-1), Shi 2023 (B1-1), Matthies 2024 (B1-2)	✓ Good (criteria), ⚠ Weak (imaging modality)
Tuberous sclerosis cutaneous features	Clinical image recognition	Portocarrero 2018 (A-20) — 2012 International Consensus criteria review	✓ Good
Gorlin syndrome / BCNS (multiple BCCs)	BCC detection criteria	Kosmidis 2023 (B2-4 excluded), B3 context refs	⚠ Weak — case report excluded; BCC detection literature (B3) supports indirectly

---

Identified Gaps and Supplementary Search Recommendations

Six conditions remain without sufficient Pillar 1 VCA literature from the current searches. The original four from block-level analysis are supplemented by pemphigus/BP (where Hocke 2023 has been recategorised to Group B, leaving no Group A anchor) and cutaneous vasculitis (where Gungor 2024 has been recategorised to Group B as a fundus-imaging study):

Gap 1: Epidermolysis bullosa (EB)

CLOSED by supplementary search S2 (2026-04-20). Three Group A VCA anchors:

• A-17 Has et al. 2020 (Br J Dermatol) — International consensus reclassification of inherited EB — 19/21
• A-18 Pfendner & Lucky 2018 (GeneReviews NBK1125) — JEB clinical features — 16/21
• A-19 Lucky et al. 2025 (GeneReviews NBK1304) — DEB clinical features — 16/21

Gap 2: Tuberous sclerosis complex (TSC) cutaneous features

CLOSED by supplementary search S3 (2026-04-20). Primary Group A VCA anchor:

• A-20 Portocarrero et al. 2018 (An Bras Dermatol) — TSC review based on 2012 International Consensus revised diagnostic criteria — 16/21. TSC cutaneous features (angiofibromas, hypomelanotic macules, shagreen patch, ungual fibromas) are formal major diagnostic criteria established by international consensus and visible in clinical photographs.

Gap 3: Dermatomyositis cutaneous features

CLOSED by supplementary search S4 (2026-04-20). Four Group A VCA anchors, including the strongest single paper in the entire supplementary search set:

• A-21 Lundberg et al. 2017 (Ann Rheum Dis) — 2017 EULAR/ACR IIM classification criteria with validated probability weights for cutaneous DM features — 20/21
• A-22 Ali et al. 2025 (Semin Immunopathol) — DM cutaneous features focus — 17/21
• A-23 Didona et al. 2023 (Ital J Dermatol Venerol) — DM clinical review — 15/21
• A-24 Chong & Werth 2022 (J Invest Dermatol) — CDASI visual scoring — 15/21

Gap 4: Morphea / localised scleroderma — image recognition

PARTIALLY CLOSED by supplementary search S5 (2026-04-20). The VCA argument rests on the clinical recognition standard (indurated ivory-white/violaceous plaques with lilac ring are image-recognisable features described in standard clinical literature), not on dermoscopy-specific criteria. The S5 search (39 results, restricted to reviews/guidelines) did not return dermoscopy-specific papers; these exist as small observational studies not captured by the filter.

• A-27 Mertens et al. 2017 (Am J Clin Dermatol) — morphea and EF clinical update — 15/21
• Residual: a dermoscopy-specific morphea search without publication-type filter is recommended for the SotA appendix in R-TF-015-011 (not load-bearing for the Pillar 1 VCA argument).

Gap 5: Pemphigus and bullous pemphigoid — clinical consensus criteria

CLOSED by supplementary search S1 (2026-04-20). Four load-bearing Group A VCA anchors identified and appraised from full text:

• A-12 Hertl 2015 (EDF/EADV S2 pemphigus guideline, JEADV) — 19/21
• A-13 Borradori 2022 (EADV S2K BP updated guidelines, JEADV) — 19/21
• A-15 Venning 2012 (BAD BP guidelines, Br J Dermatol, NHS Evidence accredited) — 19/21
• A-14 Schmidt 2021 (EADV S3 MMP guidelines, JEADV) — 17/21
• A-16 Murrell 2008 (International Pemphigus consensus definitions, JAAD) — 12/21 (context only)

Gap 6 (former Gap 5): Cutaneous vasculitis skin-image recognition

PARTIALLY CLOSED by supplementary search S6 (2026-04-20). Two Group A VCA anchors establish palpable purpura and cutaneous vasculitis skin patterns as image-recognisable clinical features that trigger the histopathological workup:

• A-25 Fraticelli et al. 2021 (Intern Emerg Med) — LCV diagnosis; "leading clinical presentation of LCV is palpable purpura" — 15/21
• A-26 Chen et al. 2024 (J Dermatol) — cutaneous vasculitis patterns in AIDs (purpura, livedo racemosa) — 16/21
• Residual: dermoscopy-specific vasculitis evidence (dermal vascular structures in purpuric lesions) was not captured by the filter and is available as small observational studies; recommended for SotA but not load-bearing for Pillar 1 VCA.

---

How the Literature Contributes to the Triangulation Argument

The 22 Group A load-bearing VCA anchors and 21 Group B SotA context papers contribute to A.2.C5 resolution at two distinct levels:

Pillar 1 VCA — what the Group A papers establish

For autoimmune dermatoses (Gap 4, approximately 3% of legacy PMS):

Image-based clinical recognition of lichen planus, discoid lupus erythematosus, and nail lichen planus is an accepted clinical standard, evidenced by:

• International multi-centre dermoscopy criteria reviews and multicenter studies (Fathy 2022; Lim 2021; Tordjman 2025 SR 12,743 cases; Sadhukhan 2025). Note: Bazzacco 2024 demoted to supporting context (14/21) after full-text review confirmed scope is neoplastic-vs-inflammatory differential, not a standalone DLE/LP criteria review. Jha 2022 excluded — full text not publicly available.
• These sources collectively show that dermoscopy is practiced, described, and validated for these conditions across multiple independent groups using clinical-photography and dermoscopy — which is the device's input modality.

Autoimmune blistering diseases (pemphigus, BP, MMP): four Group A VCA anchors confirmed from supplementary search S1. The EADV S2 pemphigus guideline (Hertl 2015), updated EADV S2K BP guideline (Borradori 2022), BAD BP guideline (Venning 2012), and EADV S3 MMP guideline (Schmidt 2021) collectively establish that the clinical visual presentation of these conditions — flaccid blisters and erosions for pemphigus; tense blisters on urticarial base for BP; scarring mucosal disease for MMP — is the recognized basis for initial clinical suspicion, and that this recognition precedes and triggers the confirmatory DIF/serology workup. The device operates at this initial-recognition step.

For genodermatoses (Gap 5, approximately 1% of legacy PMS):

• NF1 cutaneous recognition criteria: the I-NF-DC international consensus (Legius 2021, cited by 331 papers) defines café-au-lait macules, axillary/inguinal freckling, and cutaneous neurofibromas as image-recognisable primary diagnostic criteria.
• Ichthyosis phenotyping: Sun et al. (JAMA Dermatol 2021, n=1000 kindreds, PMID 34851365) demonstrates standardised clinical photographs establish genotype-phenotype associations across 32 ichthyosis genes.
• Epidermolysis bullosa: Has et al. (Br J Dermatol 2020, international consensus, 292 citations) classifies inherited EB types by clinical morphology visible in photographs; GeneReviews chapters for JEB and DEB confirm clinical features as the recognition basis.
• Tuberous sclerosis: TSC cutaneous features (angiofibromas, ash-leaf spots, shagreen patch, ungual fibromas) are formal major criteria per the 2012 International TSC Consensus (Portocarrero 2018 review).

SotA / feasibility context — what the Group B papers establish (not Pillar 1 VCA)

The 21 Group B papers establish that AI-based classification of these conditions has been implemented and benchmarked externally, demonstrating technological feasibility. These papers belong in R-TF-015-011 SotA section 3 (state of the art in image recognition). They support the Pillar 2 analytical-performance feasibility argument but do not themselves constitute VCA evidence. Key benchmarks for the SotA narrative: ResNet-50 accuracy 89.07% for LP (Eskandari); OLP AUC 0.948 (Vinayahalingam); 20-class CNN top-1 86.7% including BP (Mathur); IIF pattern recognition for AIBD (Hocke); AI severity assessment SR (Cai).

Missing surrogate-to-outcome anchor (critical gap for CER propagation)

This appraisal establishes the first half of the Pillar 1 VCA link — clinical recognition is an accepted standard — but not the second half: that accurate image-based recognition of these conditions is causally linked to improved patient outcomes.

For a Class IIb indirect-benefit device, the VCA must anchor the surrogate endpoint (accurate image recognition) to a patient outcome (e.g., reduced diagnostic delay for NF1 enabling earlier surveillance; correct pemphigus recognition preventing mucosal morbidity from delayed immunosuppression; correct ichthyosis recognition enabling genetic counselling). None of the 32 reviewed papers does this work directly.

The surrogate-to-outcome literature for these conditions must be collected separately. This is the scope of task-3b6-surrogate-endpoint-literature-review (seeded in commit d92594a51). Before the CER §Sufficiency determination borrows from this appraisal, task-3b6 must supply the outcome-linkage references; these two tasks are load-bearing in combination.

Pillar-separation note (CER author guidance)

This Pillar 1 VCA literature package is one of three evidence sources the CER will combine for the autoimmune + genodermatoses sufficiency argument. It is distinct from and complementary to:

1. MAN_2025 (Pillar 3 §4.4 Rank 11 MRMC, FP V–VI reader study) — this is supporting Pillar 3 evidence, not a Pillar 1 VCA anchor. The Tordjman 2025 SR (Group A, nail LP in FP IV–VI) is the Pillar 1 cross-reference for the dark-phototype representativeness argument; MAN_2025 provides the Pillar 3 §4.4 reinforcement.
2. R-TF-015-012 (Pillar 3 Rank 4 legacy real-world equivalent-device evidence) — this provides clinician-and-device performance data on the predecessor device, not clinical-standard VCA literature.

The three sources sit at different pillars and ranks by design. This document claims only the Pillar 1 anchors; Pillar 2 and Pillar 3 evidence lives in the respective ingredients of the triangulation strategy.

Pillar 1 VCA claim statement

The finalised tiered VCA Claim Statement is in evidence-package/pillar-1-literature.md §VCA Claim Statement. Use that version for the CER and SotA appendix. The draft below is superseded.

Gap status summary:

• Gaps 1 (EB), 2 (TSC), 3 (DM), and 5 (pemphigus/BP): fully closed — load-bearing Group A anchors confirmed.
• Gaps 4 (morphea) and 6 (cutaneous vasculitis): partially closed at clinical-recognition level (Tier 2 in pillar-1-literature.md); dermoscopy-specific evidence not captured by search filters — flagged as SotA supplement for R-TF-015-011.
• Oral LP: residual gap (Tier 3) — Jha 2022 excluded (full text unavailable); no load-bearing anchor retained. Does not materially weaken the overall VCA argument.

---

Conclusions

Pillar 1 VCA verdict (Work-stream 3)

Work-stream 3 is complete. The literature review has produced sufficient Pillar 1 Valid Clinical Association evidence for the autoimmune dermatoses and genodermatoses sub-indications to support a multi-source triangulation argument under MDCG 2020-6 §6.3.

The verdict by tier:

• Tier 1 (12 condition groups — fully closed VCA): discoid lupus erythematosus, lichen planus (skin), nail lichen planus, dermatomyositis, pemphigus group, bullous pemphigoid, mucous membrane pemphigoid, ichthyoses, neurofibromatosis type 1, neurofibromatosis type 2, tuberous sclerosis complex, epidermolysis bullosa. 22 load-bearing Group A VCA anchors confirmed from full text.
• Tier 2 (2 condition groups — partially closed): morphea and cutaneous vasculitis. Clinical-recognition standard established at the 15/21 threshold; dermoscopy-specific evidence not captured by the search filters. Not load-bearing for the VCA argument at clinical-photography level; flagged for SotA supplement only.
• Tier 3 (4 condition groups — residual gaps): oral lichen planus (mucosal), Netherton syndrome, monilethrix, Gorlin syndrome. No load-bearing anchors retrieved. These conditions represent a small minority of presentations within the two sub-indications and do not materially weaken the overall argument.

How this literature validates the device

Under MDCG 2020-1 §4.2, Pillar 1 requires that the scientific link between the device's input modality (clinical and dermoscopic images) and the clinical management decision it supports be corroborated by established medical literature — that is, that image-based clinical recognition of the target conditions is an accepted practice with known diagnostic standards.

The 22 load-bearing Group A anchors establish this link across the 14 Tier 1 and Tier 2 condition groups. Every one of these conditions has a clinically accepted first step — clinical inspection and/or dermoscopy — that triggers a confirmatory workup (histopathology, direct immunofluorescence, serology, or molecular genetic testing). The device's function maps precisely to this first step: it presents a Top-5 prioritised differential-diagnosis view to the healthcare professional, operating at the initial recognition level. It does not perform and does not claim to perform the confirmatory tests.

The literature therefore validates the device's intended function for these conditions — not by demonstrating that the device itself achieves a certain sensitivity or specificity (that is Pillar 2), but by establishing that the task the device assists with (image-based clinical recognition) is the medically accepted standard for initiating the diagnostic pathway. This is the correct and sufficient contribution of Pillar 1 evidence under MDCG 2020-1 §4.2.

What this review does not establish

This review answers the first of two questions required for the full Pillar 1 VCA of a Class IIb indirect-benefit device:

Q1 (answered here): Is image-based recognition of these conditions an accepted clinical standard? Yes — for 14 of 18 named condition groups, with load-bearing anchors confirmed from full text.

Q2 (not answered here): Is accurate image-based recognition (the device's output) causally linked to a patient-level clinical benefit (the claimed indirect benefit)? This requires the surrogate-endpoint-to-outcome literature — that earlier detection or differential ranking leads to reduced diagnostic delay, fewer complications, or improved treatment initiation. This is the scope of task-3b6-surrogate-endpoint-literature-review. Both Q1 and Q2 must be answered before the full Pillar 1 VCA argument in CER §Sufficiency can be closed.

Regulatory impact on A.2.C5

This review directly addresses the two pre-certification gaps identified in finding A.2.C5, replacing "passive surveillance" as the sole evidence source:

Gap 4 (autoimmune diseases, ~3% of use cases): Pillar 1 VCA is established. Eight autoimmune condition groups have load-bearing anchors. Oral LP is the only residual Tier 3 gap among the autoimmune sub-indication — one condition within a ~3% sub-population — and does not undermine the overall §6.5(e) argument.

Gap 5 (genodermatoses, ~1% of use cases): Pillar 1 VCA is established. Five genodermatoses condition groups have load-bearing anchors covering the major subtypes (ichthyoses, EB, NF1, NF2, TSC). Netherton syndrome, monilethrix, and Gorlin syndrome are residual Tier 3 gaps within an already rare (~1%) sub-population.

Ingredient 1 (this review) is ready to be combined with Ingredients 2–5 of the triangulation strategy. In the §6.5(e) four-test rewrite (Work-stream 5), this review serves as the Pillar 1 VCA evidence for test 3 ("is there adequate residual evidence for the claim at these sub-categories?"). Evidence for the remaining tests comes from Work-streams 1, 2, 4, and the PMCF specification.

Fitness of the device claim

No paper in this review establishes that image-based recognition alone (without histopathological, immunofluorescence, serological, or molecular confirmation) is sufficient for definitive diagnosis of any condition in scope. This is consistent with the intended device claim: probability-ranked differential diagnosis support, not standalone definitive diagnosis. The literature confirms that the step the device assists with is clinically meaningful and the recognised first step of the diagnostic pathway — not the last.

---

Recommended Next Steps

Completed (Work-stream 3)

1. Execute supplementary searches S2–S6 — COMPLETE (2026-04-20). All six condition-group gaps addressed via primary blocks A–D and supplementary searches S1–S6.
2. Obtain full texts for all load-bearing Group A papers — COMPLETE (2026-04-20). 26 manuscripts in included-manuscripts/. Jha 2022 (A-07) excluded — full text not publicly available; oral LP is Tier 3 residual gap. Final load-bearing count: 22 papers.
3. Write evidence-package/pillar-1-literature.md — COMPLETE (2026-04-20). 26-reference list (15 load-bearing autoimmune + 7 load-bearing genodermatoses + 4 supporting context). Tiered VCA Claim Statement finalised. Reviewed by BSI clinical auditor and Celine/Saray pillar-mapping agents; two revision passes completed.

Active next steps (can run in parallel)

4. Complete task-3b6 — surrogate-endpoint literature review (../task-3b6-surrogate-endpoint-literature-review/). This is a hard prerequisite for Work-stream 5: the CER §Sufficiency section cannot be written until task-3b6 establishes the surrogate-to-patient-outcome linkage for the indirect-benefit claim.
5. Execute Work-stream 1 — Pillar 2 algorithm V&V (see CLAUDE.md §Work-stream 1). Query the curated labelling dataset for autoimmune and genodermatoses examples; compute per-category sensitivity, specificity, AUC, Top-1, Top-5. Produce evidence-package/pillar-2-algorithm-vv.md, cross-referencing R-TF-028-*.
6. Execute Work-stream 2 — Rank 7 legacy PMS filter (see CLAUDE.md §Work-stream 2). Query the legacy 250,000+ report corpus for autoimmune/genodermatoses keywords and ICD-11 codes; apply rule-of-three upper bounds where zero events. Produce evidence-package/rank-7-legacy-pms-filter.md, cross-referencing R-TF-007-003.
7. Execute Work-stream 4 — Fast focused MRMC (see CLAUDE.md §Work-stream 4). This is a separate study from MAN_2025 (different sub-indication and image set). Pre-specify protocol (≥30 images per category group, ≥5 readers, pre-specified acceptance criteria). Produce evidence-package/mrmc-protocol.md and evidence-package/mrmc-results.md.

Blocked until steps 4–7 are complete

8. Execute Work-stream 5 — CEP/CER edits + PMCF spec + claim language (see CLAUDE.md §Work-stream 5). Requires WS1, WS2, WS4, and task-3b6 all complete. Produce four-test-rewrite.md, pmcf-activity-spec.md, narrowed-claim-language.md. Propagate into CEP lines 900–908, CER §Representativeness and §Sufficiency, PMCF Plan R-TF-007-002, Intended Purpose reusable, and IFU.

Housekeeping (before finalising R-TF-015-011 SotA appendix)

9. Append evidence-package/pillar-1-literature.md to R-TF-015-011 State of the Art as a new named section. This is a WS5 sub-task but can be done independently once WS3 is complete.
10. Deduplicate Group B SotA papers before incorporating them into R-TF-015-011: Hocke 2023 appears in blocks A1 and C1; Vinayahalingam 2024 appears in blocks A1 and A3. Count each once in the final SotA reference list.