Can AI help skin cancer pathways maintain high sensitivity

A newly published study in Skin Health and Disease evaluated DermDx, MetaOptima’s AI-powered skin lesion assessment and triage product, in a blinded, real-world NHS setting.

DermDx analyzes dermoscopic images and produces a calibrated malignancy risk score, together with a low- or high-risk triage classification. The system is designed to support clinical decision-making across commonly used dermoscopy devices and image-capture hardware.

In the AI-SCSS study, researchers evaluated DermDx on 1,024 lesions referred through a suspected skin cancer pathway.

Key findings:

• DermDx identified 296 of 298 confirmed skin cancers, achieving 99.3% sensitivity.
• Sensitivity across malignant and premalignant lesions was 98.1%.
• The algorithm classified 26.7% of lesions as low risk, indicating that face-to-face consultations could have been reduced by the same proportion within the study cohort.
• The area under the receiver operating characteristic curve was 0.918 for malignant and premalignant lesions.
  
  

Developed by MetaOptima, DermDx combines three complementary AI vision models and was trained using labelled dermoscopic images. Its intended role is not simply to classify images, but to help clinical teams prioritize higher-risk lesions and use specialist capacity more effectively.

The results add valuable real-world evidence to a field that has relied heavily on controlled image-reader studies. They also point toward a future in which AI and dermatology expertise work together to improve patient pathways.

Congratulations to our team, our partners at ISLA and NHS and clinical teams behind the AI-SCSS study.

Read the open-access paper: https://doi.org/10.1093/skinhd/vzag122