To evaluate the diagnostic performance of dermoscopy alone versus dermoscopy supported by AI or AI alone in melanoma detection through a systematic review and meta-analysis of prospective studies, highlighting AI's potential to enhance diagnostic accuracy.
Key Findings:
AI systems demonstrated diagnostic performances comparable to or exceeding those of expert dermatologists, with specific metrics indicating sensitivity and specificity improvements.
Prospective studies provide a more reliable assessment of AI's diagnostic capabilities compared to retrospective studies.
The integration of AI in clinical practice for melanoma detection is promising but requires further validation in real-world settings.
Interpretation:
The findings suggest that AI can enhance melanoma detection accuracy, but the clinical readiness for its widespread implementation needs further investigation, particularly in real-world applications.
Limitations:
Most studies evaluated were limited in scope and may not fully represent real-world clinical scenarios, with potential biases in study designs and data collection methods affecting generalizability.
Conclusion:
AI shows potential as a decision-support tool in melanoma diagnostics, but further prospective validation is essential for clinical integration, emphasizing the need for additional studies.
by Sara Laiouar-Pedari, Arlene Kühn, Christoph Wies, Carina Nogueira Garcia, Jana Therés Winterstein, Lukas Heinlein, Annemarie Hoffsommer, Tirtha Chanda, Sarah Haggenmüller, Titus J. Brinker
