Rapid advances in artificial intelligence have established medical image analysis as a cornerstone of intelligent healthcare. Deep learning techniques, including convolutional neural networks (CNNs), graph convolutional networks (GCNs), multi-layer perceptrons (MLPs), and vision transformers (ViTs) architectures, substantially enhance performance in medical image classification and segmentation. This progress advances diagnostic accuracy, robustness, and efficiency. This book systematically surveys deep learning models for medical image analysis. It documents the evolution from MLPs and CNNs to hybrid attention architectures, with technical analysis of 9 recent methodologies. Core topics cover: multi-scale feature fusion, multi-branch CNN structures, graph-based feature modeling, region-aware attention mechanisms, adaptive positioning modules, and lightweight model design. This book addresses: (1) MLP-based models for disease classification; (2) integrated CNN and ViT approaches for spatially contextualized learning; (3) GCNs for topological and relational representation; and (4) lightweight models for efficient deployment under resource constraints. Each chapter examines representative publications, summarizing methodological innovations, architectures, experimental results. This work integrates theory with implementation, serving as a reference for researchers and professionals in medical imaging, computer-aided diagnosis, and biomedical AI. It establishes foundations for current deep learning paradigms and future development.
