National Health Service hospital trusts in the United Kingdom are deploying an autonomous artificial intelligence system named DERM to evaluate suspicious skin lesions, helping reduce unnecessary biopsies and increasing patient intake capacity by 62 percent as dermatology waiting times face demands.
Urgent suspected skin cancer referrals in England have nearly tripled since 2009, according to data highlighted at the European Academy of Dermatology and Venereology (EADV) Congress 2026, while only about 6 percent of those cases result in a true malignant diagnosis. With roughly one in four dermatologist roles remaining unfilled across the United Kingdom, hospital clinics have increasingly turned to autonomous medical technology to triage patients and streamline clinical workflows.
Real-World Deployment at St. Luke’s Hospital and Chelsea and Westminster Hospital
The AI system DERM, developed by the U.K.-based company Skin Analytics, operates by analyzing clinical and dermoscopic smartphone images within minutes. Medical staff capture three images of a suspicious mole or lesion using a dermatoscope attachment, allowing the algorithm to assess cancer risk before a doctor reviews the case. At St. Luke’s Hospital in Bradford, the technology was launched on April 30, making Bradford Teaching Hospitals the first NHS trust in West Yorkshire to adopt the system. Across its roughly 5,000 annual skin cancer referrals, the clinic increased patient examinations from 24 to 32 per session following the rollout, while achieving a 10 percent drop in unnecessary biopsies.
“This has had a big impact on all the bottlenecks – right from waiting times for patients to unnecessary investigations being done and to being seen by a clinician.”
Zakir Sharif, plastic surgeon and clinical lead for skin cancer at Bradford
Similar deployments at Chelsea and Westminster Hospital and other participating trusts have demonstrated that the software can autonomously discharge low-risk cases without initial clinician review. Data presented from a 16-month study of 8,391 patients across two hospital sites showed that the device saved approximately 2,851 clinician hours, translating to roughly 8,553 additional face-to-face dermatology appointments created during the study window.
Patient Impact and Clinical Perspectives on Low-Risk Triage
For patients waiting for urgent referrals, the accelerated triage provides rapid reassurance. Lawrence Patten, a 73-year-old former builder from Bradford with a history of multiple myeloma and previous squamous cell carcinoma on his right forearm, was urgently referred by his general practitioner for a suspicious mole on his back. Although the AI flagged the lesion as suspicious, a subsequent physical examination by Dr Nader Ghaderi, associate specialist in dermatology, indicated it was likely benign, though it was scheduled for removal and testing to confirm.
“It’s very good news. It’s not ticking away in my mind, which is one of the biggest things with cancer. The sooner they find things, the quicker they can sort it.”
Lawrence Patten, Patient
Clinicians emphasize that the primary value of the software lies in reallocating scarce medical expertise. Dr. Thomas, consultant dermatologist at Chelsea and Westminster Hospital, and honorary clinical lecturer at Imperial College London, noted that time saved from evaluating routine lesions can be redirected toward patients with aggressive skin cancers or severe inflammatory skin diseases.
Autonomous Pathway Shows High Sensitivity but Six False Negatives
Evaluating safety across a national dataset encompassing the study sites revealed that the autonomous pathway achieved a sensitivity exceeding 98 percent for invasive melanoma, squamous cell carcinoma, and basal cell carcinoma, alongside a specificity of 72.1 percent. Post-market surveillance identified six false-negative cases that had been autonomously discharged through the pathway—comprising five basal cell carcinomas and one melanoma in situ. These instances were subsequently identified and reviewed.
To maintain safety standards, researchers stress that continuous oversight remains mandatory. Dr. Thomas pointed out that deploying automated diagnostic tools requires ongoing monitoring, learning from misclassifications, and ensuring patients remain informed about warning signs. Meanwhile, the National Institute for Health and Care Excellence has recommended DERM for a three-year evidence-gathering evaluation across more than 40 NHS trusts nationwide.
Researchers Combine Raman Spectroscopy with Machine Learning Models
While hospital trusts scale software evaluations, independent research groups continue exploring optical and molecular methods for non-invasive detection. Investigators at Florida Atlantic University published preliminary laboratory findings combining Raman spectroscopy with machine learning models. Researchers analyzed spectra from tissue samples, including normal skin, basal cell carcinoma, and squamous cell carcinoma.
