Comprehensive Summary
This study evaluated the reliability of an AI-based dermatology device in labeling benign skin lesions in comparison to teledermatologists. The authors analyzed over 2000 lesions assessed across 13 NHS sites, comparing how AI classified them versus how a second read teledermatologist versus a hospital teledermatologist. Among 2133 lesions, the AI device showed high predictive accuracy for most benign categories. Second read teledermatologists had a similar performance, with no statistically significant differences across subcategories. The authors highlight that AI’s reliability to correctly predict benign skin lesions is very similar to how specialists do, indicating their potential for clinical implementation.
Outcomes and Implications
Benign lesions are a large part of what dermatologists look at, so this technology could improve patient flow and provide reliable reassurance of accurate diagnoses. It could also be important in areas who don’t have access to specialist care immediately, or have limited access. The authors argue that with continued validation and monitoring, the AI system could help assess benign lesions across diverse populations and areas.