Tejal Somvanshi
An autonomous AI system deployed across two UK hospitals triaged 8,391 skin lesion patients over 16 months — and researchers calculated that the specialist time saved could support the equivalent of more than 8,500 additional face-to-face dermatology appointments. The study was presented at the European Academy of Dermatology and Venereology (EADV) conference in October 2026.
The 8,500 appointments figure is an estimated capacity equivalent, not 8,500 appointments already delivered. The distinction matters. What the study demonstrates is that by using autonomous AI to assess and discharge lower-risk patients before they reach a consultant's desk, the system can redirect specialist time toward patients with genuine suspected malignancies.
For patients enduring weeks of anxiety waiting for a specialist skin check, autonomous AI screening offers a real bottleneck breaker. By identifying lower-risk cases at local clinics, these diagnostic models free up specialist clinical capacity for patients with genuine suspected melanomas — while maintaining consultant oversight for higher-risk cases.
The algorithm was developed using deep learning foundation models. Its primary application is not replacing biopsies. It is autonomously discharging a subset of patients classified as lower-risk and referring higher-risk cases to specialist clinicians, so human dermatologists can direct their clinic time toward the patients most likely to need them.
Related AI-health coverage includes Karmactive’s report on New AI Spots Skin Cancer: PanDerm.
How Safe Is AI Triage for Skin Cancer?
The study reported six false-negative discharges — cases where the AI cleared a patient who was later identified through follow-up surveillance as requiring further assessment. All six were detected through surveillance mechanisms without adverse outcomes being recorded in the available follow-up period.
That number is more informative than the headline appointment figure. The AI did not have a perfect record. But the surveillance system caught the errors, and no documented harm resulted in the reported follow-up. This is the safety architecture that underpins this type of clinical pathway: the algorithm handles the high-volume, lower-risk end of the queue while human oversight layers catch edge cases.
The reported study evaluated one autonomous AI-supported pathway in two UK hospitals. Its findings do not establish that all dermatology AI systems perform at consultant level or that the pathway is suitable for every patient or NHS setting.
For melanoma treatment context, see the Moderna-Merck personalised cancer vaccine / melanoma.
The NHS is operating under sustained two-week-wait pressure across multiple specialties, and dermatology is among the highest-volume referral categories. Scaling autonomous pre-triage nationally would reduce the administrative burden of routing every lower-risk mole referral through the same bottleneck as suspected melanomas.
A formal NICE technology appraisal guideline covering autonomous AI dermatology triage should be checked directly against the NICE website for its current status.
