Hiring in life sciences? Share your open positions with our professional community. Read more Close

Advertisement

Artificial Intelligence in Skin Cancer Detection: Is Clinical Deployment Outpacing Experimental Validation?

Created on 14 Nov 2026

Authors

D Reifs, L Casanova, F Hawke, A Fellas, A Eidels, E Stevenson, E Croft, X Girones, S Diez-Alvarez, D Santos, A Coda

Published in

Experimental dermatology. Volume 35. Issue 11.

Abstract

Deep neural networks classify dermoscopic and clinical images of skin lesions with accuracy that matches, and in curated comparisons exceeds, experienced dermatologists. Regulators have responded: an elastic scattering spectroscopy device with an embedded classifier is authorised for primary care in the United States, an image-based system has conditional approval in the National Health Service, and consumer applications are marketed directly to the public. In this Viewpoint we argue that clinical deployment is advancing faster than the experimental evidence that investigative dermatology would normally demand. Performance measured on internal test sets falls when algorithms are exposed to external images, non-dermoscopic photographs or other populations. Networks learn confounders, such as surgical ink markings, as readily as morphology, and the histopathological reference standard is itself subject to substantial observer variation. Public training datasets record skin phototype for about 2% of images and contain almost no darkest-phototype, acral or nail images. Regulated devices with prospective evidence achieve high sensitivity or negative predictive value at the cost of low specificity, shifting referral and biopsy rates with unmeasured effects on patient outcomes and overdiagnosis, and the single randomised trial to date improved diagnostic accuracy, chiefly among non-dermatologists, without measuring those outcomes. We identify where the largest gains lie, in primary care, allied health and under-served regions, and propose an experimental agenda: phototype-stratified external validation, shortcut-detection and counterfactual explanation studies, consensus histopathological labelling, pragmatic randomised trials with outcome endpoints reported to CONSORT-AI, and post-market surveillance under the European Medical Device Regulation and Artificial Intelligence Act.

PMID:
43031259
Bibliographic data and abstract were imported from PubMed on 14 Nov 2026.

Read full publication at:
Please sign in to see all details.

Advertisement

Stats

  • Community rating n/a 0 votes
  • Reviewers' rating n/a 0 votes
  • Your rating

1-terrible, 9-excellent. How would you rate this publication? Sign in in to submit your rating.

  • Recommendations n/a n/a positive of 0 vote(s)
  • Views 10
  • Comments 0

Recommended by

  • No recommendations yet.

Post a comment

You need to be signed in to post comments. You can sign in here.

Comments

There are no comments yet.

Advertisement