A Comprehensive Review Tracing the Evolution of Volumetric Medical Imaging Analysis from Classic CNNs to Emerging AI-Agents

Volumetric medical imaging analysis redefines modern healthcare by allowing for early diagnosis, personalized prognosis, and precision-guided treatment planning. During the past decade, the field has transitioned from handmade features to deep learning systems capable of uncovering rich volumetric biomarkers and leveraging heterogeneous datasets for predictive and generative modeling. In this review, we trace this transformative journey through four paradigms. First, we revisit classic deep architectures, CNNs, RNNs, and early transformers, which laid the foundations of volumetric learning. Second, we examine generative models, from VAEs and GANs to diffusion approaches that advanced synthesis, augmentation, and reconstruction. Finally, we highlight the rise of foundation models that employ vision language models, large language models, and AI-agents that go beyond static models, orchestrating reasoning, decision making, and task adaptation in dynamic clinical workflows. For each paradigm, we critically assess methodological innovations, strengths, limitations, and comparative performance in segmentation, classification, detection, reconstruction, and report generation. Beyond synthesizing progress, we identify persistent challenges, including data scarcity, generalization between institutions, and clinical trustworthiness, and outline emerging frontiers in multimodal fusion, explainable AI, and human-AI collaboration. To support further research, we provide a GitHub repository that includes popular 3D medical imaging datasets along with recent 3D models in our shared GitHub repository.