Multimodal artificial intelligence for enhanced skin cancer diagnosis and prognosis

Skin cancer represents one of the most common and clinically significant malignancies worldwide, with rising incidence and substantial mortality, particularly melanoma, which accounts for approximately 75–80% of skin cancer-related deaths despite comprising < 5% of cases. Early detection and accurate prognostication remain essential for improving patient outcomes, yet subjectivity, resource constraints, and inter-observer variability often limit conventional diagnostic methods. Recent advances in artificial intelligence (AI), particularly deep learning, have demonstrated remarkable potential in enhancing diagnostic precision, risk stratification, and personalized treatment planning. Convolutional neural networks, vision transformers, and hybrid architectures have achieved dermatologist-level accuracy in classifying dermoscopic and clinical images, while multimodal approaches integrating histopathology, genomics, and patient metadata have improved prognostic assessments. Despite these advances, significant challenges remain, including dataset biases, lack of interpretability, regulatory barriers, and disparities in access to AI technologies. Addressing these challenges through explainable AI, federated learning, robust clinical validation, and equitable deployment strategies will be crucial for translation into routine practice. This review synthesizes current evidence on AI applications in skin cancer diagnosis and prognosis, highlights key technological innovations, and outlines the challenges and future directions necessary to ensure safe, effective, and globally accessible integration of AI into dermatological care.​.