[Computer-assisted skin cancer diagnosis : Is it time for artificial intelligence in clinical practice?]

Area of Science:

• Dermatology
• Medical Imaging
• Artificial Intelligence

Background:

• Artificial intelligence (AI) is increasingly utilized in medicine, particularly for image-based diagnosis.
• AI demonstrates significant potential in skin cancer detection, yet a gap exists between its perceived and actual clinical relevance.

Purpose of the Study:

• To summarize study findings on AI-driven skin cancer diagnosis using computer-based systems.
• To discuss the significance of these findings for daily dermatological practice, focusing on dermoscopic images.

Main Methods:

• A selective literature search was conducted on recent trials.
• Included studies employed machine learning, specifically convolutional neural networks (CNNs), for image data classification.

Main Results:

• Computer algorithms achieved high precision in detecting pigmented and nonpigmented skin neoplasms, comparable to dermatologists.
• The combination of physician assessment and AI yielded the best diagnostic results.
• AI systems are generally accepted by patients and physicians.

Conclusions:

• AI diagnostic systems are not yet applicable in daily practice due to experimental testing and unclear classification criteria.
• Lack of transparency in AI decision-making needs to be addressed.
• Clinical studies are essential to validate the applicability of AI assistance systems in routine dermatology.