Dual Pathways in Dermatological AI: A Machine Learning Survey on Image-based and Biopsy- Driven Skin Disease Classification

Skin diseases represent a significant global health burden, requiring timely and accurate diagnosis to prevent disease progression and reduce clinical workload. Recent advances in artificial intelligence (AI) and machine learning (ML) have enabled automated diagnostic systems that analyze dermatological images and biopsy-derived data with increasing accuracy. However, existing surveys often remain descriptive, lack of methodological synthesis, and underexplore clinical translation challenges. This paper presents a comprehensive and critical survey of AI-based skin disease classification systems developed between 2020 and 2025, with a dual focus on image-based and biopsy-driven approaches. A structured taxonomy is introduced to categorize existing methods based on data modality, learning paradigm, model architecture, and clinical objectives. Beyond performance reporting, this work provides analytical insights into why certain models succeed under specific conditions and where their limitations arise. The survey further synthesizes reported results through a comparative quantitative analysis, highlighting trends in diagnostic accuracy, generalizability, and scalability. Key challenges related to dataset bias, annotation quality, explainability, clinical integration, and regulatory compliance are examined in detail. Finally, a prioritized research roadmap is proposed to guide future developments toward trustworthy, multimodal, and clinically deployable dermatological AI systems.