Comprehensive Summary
The article discusses the ability of artificial intelligence to classify skin cancer compared to that of clinicians. Salinas et al. screened 3 separate databases for studies including statistics of sensitivity and specificity for both AI screening models and clinicians. When compared against both general and expert clinicians, AI models had higher specificity and sensitivity. However, the difference between specificity and sensitivity scores of the two groups was higher when comparing AI models to generalists than it was when compared to expert clinicians, indicating that it could perform similarly to general clinicians in a preliminary analysis of skin conditions, but further analysis would be more reliably done by expert clinicians.
Outcomes and Implications
This research is important as it expands opportunity for medical analysis of skin conditions, especially to groups which lack access or funding for professional analysis. While AI models cannot be used to definitively diagnose conditions and should not suggest treatment methods, they could be used to help differentiate between common harmless skin abnormalities and abnormalities which require medical attention. The authors suggest limiting use of AI algorithms in a medical setting and further development of AI-assistance before integration in practice.