AI Boosts Melanoma Diagnosis Accuracy to 90%

1. Artificial intelligence triage for melanoma shows more than 90 percent sensitivity and is now in trials within the National Health Service.

2. Referrals for suspected skin cancer in England have doubled in the past decade, making early triage critical.

England’s health service began piloting an artificial intelligence tool called Deep Ensemble for Recognition of Malignancy, abbreviated as DERM, across several hospitals as part of a three-year evaluation, reported on August 27 in Nature. The system analyzes skin lesion images and prioritizes which patients should be seen first, offering sensitivity for melanoma that exceeded 90 percent in early studies. This level of accuracy compares favorably with most board-certified dermatologists, who average closer to 82 to 85 percent sensitivity in large audits. Importantly, DERM is being tested on patients with a range of skin tones and lesion types, addressing concerns that earlier models did not perform equally across populations. Referrals for suspected melanoma in the National Health Service have increased from roughly 450,000 in 2013 to more than 1 million in 2023, straining capacity. In some regions, patients wait over three months for initial dermatology review, a gap that may worsen outcomes if dangerous lesions are missed. Artificial intelligence may help by consistently triaging which lesions require dermoscopy and biopsy most urgently. Clinicians emphasize that the system does not replace dermatologists but instead serves as a screening layer. Safety monitoring will track false negatives closely to ensure early cancers are not overlooked. If results hold, NHS leaders say the program could shorten waiting times by weeks. For physicians, the promise is a more efficient pathway to identify high-risk lesions. By embedding artificial intelligence into referral systems, national health services may finally balance growing demand with limited specialist supply.