Recent advances in artificial intelligence for melanoma: A review of history, models, datasets, applications, and ethical and legal considerations

Abstract

Melanoma is one of the most lethal and aggressive forms of skin cancer, often presenting as evolving pigmented lesions. This review aims to examine melanoma from a holistic perspective by evaluating integrative strategies that combine Artificial Intelligence (AI) with traditional diagnostic approaches. It explores how AI, particularly Machine Learning (ML) and Deep Learning (DL) techniques, can improve the early detection, classification, prognosis and treatment of melanoma, while addressing limitations of dermoscopy and histopathology, to enhance diagnostic accuracy, efficiency, and accessibility. A comprehensive literature review approach was conducted on recent studies and datasets related to AI applications in melanoma detection, classification, and analysis. The review focuses on ML and DL algorithms applied to analyze dermoscopic images, their performance, data requirements, and clinical relevance. ML and DL models have demonstrated high accuracy in melanoma identification and classification from dermoscopic images. When trained on large, well-annotated datasets, these models outperform traditional dermoscopic assessments, enabling faster analysis, reducing human error, and improving diagnostic consistency across clinical settings. AI-based ML and DL approaches show strong potential to support clinicians in the early and accurate detection and management of melanoma. By complementing clinical expertise, these integrative technologies can enhance diagnostic outcomes, reduce costs, and facilitate timely clinical interventions. Continued research and improvement of AI models and datasets are essential for their successful use in clinical practices. • Skin cancer is one of the most common oncological disease; melanoma is also one of the most aggressive subtype. • Early detection of melanoma is critical for survival but remains challenging due to diagnostic complexity. • Current diagnostic tools are dermoscopy and histopathology, which face limitations in accuracy and consistency. • AI and machine learning offer promising tools for early, accurate melanoma detection and prognosis. • Deep learning excels in analyzing dermoscopic images, improving lesion classification and treatment guidance.

Ähnliche Arbeiten