Key takeaways:

• An AI medical device reduced routine follow-up appointments (12% vs. 27%) and biopsies (26% to 29% vs. 43%) compared with standard care.
• Sensitivity for malignancy exceeded 97% across all skin cancer types.

An autonomous AI device was successfully integrated into a real-world health system and managed 94% of urgent suspected skin cancer referrals, according to a presentation at the European Academy of Dermatology and Venerology Congress.

The researchers also found that the AI medical device safely discharged over a quarter of patients without a dermatologist review over the course of 16 months at two hospital sites in the U.K., according to Lucy Thomas, MPharm, MBChB, FRCP, consultant dermatologist at Chelsea & Westminster Hospital in London.

Data derived from Thomas L, et al. Abstract #2145. Presented at: European Academy of Dermatology and Venerology Congress; Sept. 30-Oct. 3, 2026; Vienna.

“This translated into approximately 2,850 hours of clinician time, equivalent to more than 8,500 face-to-face dermatology appointments,” Thomas told Healio. “The findings suggest that autonomous AI has the potential to create meaningful additional specialist capacity within a service that is facing growing demand and significant workforce shortages.”

According to the presentation, the U.K.’s urgent suspected skin cancer (USSC) referrals have tripled since 2009 and continue to rise annually. However, the conversion rates remain low, hovering at about 6%.

“At the same time, around 1 in 4 dermatologist posts remain unfilled,” Thomas said. “As a result, significant specialist time is being devoted to assessing patients who do not require ongoing dermatology input, while other patients face delays in accessing care.”

Because AI has “demonstrated diagnostic performance comparable to dermatologists,” according to the presentation, researchers set out to assess the real-world deployment of autonomous AI to triage skin cancer cases.

After a pilot evaluation completed by teledermatologists, researchers implemented the AI device in December 2024 and July 2025 at two National Health Service hospital sites in the U.K. The autonomous AI device validated lesions with dermoscopic confirmation and image quality assessment, then diagnosed and classified lesions based on a risk hierarchy. Benign cases were discharged autonomously with safety-netting advice that did not require clinician review, whereas high-risk cases were triaged to a teledermatologist, according to the presentation.

A total of 8,391 patients of all skin types were managed via the AI device, with patient-consent reaching 86%. The device safely discharged 31% and 25% of benign cases at each hospital site. Teledermatologists discharged an additional 24% and 25%, according to the researchers.

The benefits of using the autonomous AI device were twofold, according to the researchers, who noted that AI reduced routine follow-up appointments compared with standard teledermatology (12% vs. 27%) and lowered the number of biopsies compared with face-to-face care (26% to 29% vs. 43%).

Sensitivity for malignancy exceeded 97% across all skin cancer types and specificity was 72.1% (95% CI, 71.7%-72.6%), according to the researchers. Five false negatives occurred for basal cell carcinoma and one for melanoma in situ.

Researchers estimated that the AI device gave clinicians a 62% increase in clinical capacity, opening the possibility for dermatologists to see more patients when necessary, according to Thomas.

“This creates an opportunity to redirect scarce specialist expertise toward patients whose outcomes depend most on rapid access to dermatological care,” Thomas told Healio. “The conversation should therefore move beyond whether AI can identify skin cancer and focus on how released capacity can be used to improve patient outcomes right across the wider dermatology service.”

For more information:

Lucy Thomas, MPharm, MBChB, FRCP, is a consultant dermatologist at Chelsea & Westminster Hospital and honorary clinical lecturer at Imperial College in London. Thomas can be reached at lucy.thomas37@nhs.net.