Dermatology · Autonomous skin-lesion triage and assessment
DERM
Skin Analytics
DERM is a genuine regulatory landmark — in February 2025 it became the first skin-cancer AI in the world to receive a Class III CE mark under the EU Medical Device Regulation authorising autonomous clinical decisions, and in the UK it is a UKCA Class IIa device conditionally recommended by NICE for early NHS use while evidence is gathered. The independent evidence is strong for a device of its class: NICE's external assessment group pooled a sensitivity of 96.1 percent for detecting any malignant lesion at a specificity of 65.4 percent, broadly comparable to dermatologists. Two honest caveats hold it at two marks rather than three. There is no FDA marketing authorisation — the company cites an FDA Breakthrough Device Designation, which is a development pathway, not a clearance — and the evidence base under-represents darker skin: the prospective DERM-003 study was 96.9 percent Fitzpatrick skin types I to III, and NICE noted the same limitation, while its analysis estimated that just under one percent of eligible lesions could be both malignant and incorrectly discharged.
Performance Metrics
Clinical Evidence
DERM's evidence base pairs manufacturer-authored prospective validation with an independent national health-technology assessment, and the two together give a fuller — and more cautious — picture than the manufacturer's headline accuracy figures alone. The pivotal manufacturer-authored prospective study is DERM-003 (Frontiers in Medicine 2023, PMID 37869160): a prospective, multicentre, single-arm, masked study across four UK NHS Trusts that recruited 572 patients providing 611 suspicious lesions, evaluating DERM's identification of squamous-cell and basal-cell carcinoma, pre-malignant and benign lesions. An earlier manufacturer-affiliated study (Phillips M, et al., JAMA Network Open 2019, PMID 31617929) reported melanoma-detection accuracy on selected lesions comparable to specialists, and a real-world post-deployment analysis of about 14,500 cases (Frontiers in Medicine 2023, PMID 38020164) reported high melanoma sensitivity. These are manufacturer studies and establish plausibility. The independent anchor is the NICE early value assessment (NIHR Health Technology Assessment, PMID 41660864; NICE guidance HTG746, published May 2025). Its external assessment group pooled a sensitivity of 96.1 percent (95% CI 95.4 to 96.8) for detecting any malignant lesion at a specificity of 65.4 percent (95% CI 64.7 to 66.1), and judged the diagnostic accuracy of autonomous DERM broadly similar to that of dermatologists. The EAG estimated that automated use could roughly halve dermatologist referrals among eligible lesions, but that just under one percent of eligible lesions would be both malignant and incorrectly discharged — the false-negative floor that defines the safety question for autonomous discharge. Two limitations bound the evidence and are load-bearing. First, skin-tone representation: DERM-003 was 96.9 percent Fitzpatrick skin types I to III, and NICE independently flagged sparse evidence in darker skin, so performance in Fitzpatrick IV to VI is not well established. Second, an independent methods critique and the manufacturer's response were published as an exchange (PMID 38379561 and PMID 38756949), and the NICE recommendation is explicitly conditional — early use while further real-world evidence is collected, not a settled endorsement.
| Study | Design | n | Sensitivity | Specificity | AUC | Published |
|---|---|---|---|---|---|---|
| DERM-003 — multicentre, single-arm, masked prospective study, 4 UK NHS Trusts (manufacturer-authored) | ProspectiveProspective | 572 | — | — | — | Frontiers in Medicine, 2023; PMID 37869160; 611 lesions; 96.9% Fitzpatrick I–III |
| NICE external assessment group pooled analysis (independent) | ProspectiveProspective | 0 | 96.1% | 65.4% | — | NIHR HTA, 2026; PMID 41660864; NICE HTG746 (2025); autonomous DERM ~ dermatologist accuracy |
| Phillips M, et al. — melanoma detection on selected lesions (manufacturer-affiliated) | RetrospectiveRetrospective | 0 | — | — | — | JAMA Network Open, 2019; PMID 31617929; accuracy comparable to specialists |
| Real-world post-deployment analysis (manufacturer-authored) | RetrospectiveRetrospective | 14,500 | high melanoma sensitivity (see paper) | — | — | Frontiers in Medicine, 2023; PMID 38020164 |
Clinical Pulse
Verified clinician sentiment
Verified clinician reviews launching soon · Apply to contribute →
Reviews are signed by clinicians verified against their regulator's public register.
Inside the Auris+ Listing
Five more sections complete this device’s Auris+ Listing.
Decision Ledger
ProPro unlocks a private, cross-vendor log of every case you read on this device — what the AI called, what you concluded, and a one-line reason if you overrode it. Ready for the EU AI Act's deployer logging obligations when they land in 2028.
Clinical Evidence Deep Dive
ProPro unlocks the structured clinical-evidence summary — study count, target patient population, and a tabular accuracy-metrics view drawn from peer-reviewed sources.
Peer-Reviewed Publications
ProPro unlocks the curated peer-reviewed publication list with PubMed cross-links — the citation backbone of every editorial verdict.
Post-Market & Regulatory Conditions
ProPro unlocks the post-market surveillance summary, recall record, and the conditions of approval that bound real-world use.
AI Algorithm Version History
ProPro unlocks the chronological record of algorithm version changes — what changed when, drawn from manufacturer changelogs and regulatory filings.
Regulatory Approvals
Safety Record
The Australian ARTG entry was cancelled at the sponsor's request — a commercial withdrawal, not a regulator-initiated safety recall.
No DERM recall or safety communication from a regulator was identified in publicly available sources as of July 2026. One registry action is easily misread and should be stated plainly: the Australian ARTG entry was cancelled at the sponsor's own request in April 2024 — a commercial withdrawal, not a safety recall. The principal clinical-safety consideration is intrinsic to the autonomous-discharge model: NICE's external assessment group estimated that just under one percent of eligible lesions could be both malignant and incorrectly discharged, so the false-negative rate — not a device malfunction — is the dominant risk, and it is the reason the UK deployment retains a second-read option. The under-representation of darker skin tones in the validation evidence is a further equity-relevant safety caveat: performance in Fitzpatrick skin types IV to VI is not well established.
Intended Use & Indications
DERM (Deep Ensemble for Recognition of Malignancy) is an AI-based skin-lesion analysis technology intended for the screening, triage and assessment of suspected skin-cancer lesions in people aged 18 and over. It is used within teledermatology services after referral from primary care: a dermoscopic image of the lesion is captured, and DERM classifies it to support a decision to discharge a benign lesion or refer a suspicious one onward. The regulatory envelope differs by jurisdiction. In the EU, DERM holds a Class III CE mark under the Medical Device Regulation that authorises fully autonomous clinical decisions — discharge without clinician review. In the UK it is a UKCA-marked Class IIa device and, under the NICE early-value-assessment recommendation, is deployed either as an automated tool or with a healthcare-professional second read while further real-world evidence is gathered. In every setting DERM triages rather than delivers a histological diagnosis; a lesion it flags is referred for specialist assessment and, where indicated, biopsy.