ARTIFICIAL INTELLIGENCE IN DERMATOLOGY: A COMPREHENSIVE REVIEW OF DIAGNOSTIC AND THERAPEUTIC APPLICATIONS, CHALLENGES AND FUTURE PERSPECTIVES

Abstract

Skin diseases are among the most common diseases affecting 1.9 billion people worldwide. They remain a major cause of non-fatal disease burden, in part because diagnosis of skin diseases is often challenging due to high visual similarity of skin conditions, subjectivity of clinical diagnosis, and lack of dermatologists worldwide. Deep learning-based artificial intelligence (AI) applied to dermatological images has emerged as a promising tool for dermatological diagnosis and treatment in the last decade. This review provides a comprehensive overview of skin diseases, conventional diagnosis and treatment in dermatology, and core AI technologies, including convolutional neural networks (CNN), computer vision, natural language processing (NLP), and explainable AI, that are currently applied in dermatology for a range of applications including skin cancer, inflammatory and pigmentary skin diseases, disease severity scoring, individualized treatment planning, teledermatology, and even dermatological drug discover. We also discuss current regulatory developments including the first AI-enabled primary-care skin cancer device authorized by US Food and Drug Administration in 2024.