Foundations of Intelligent MRI: From Spin Physics to AI Models

This book develops a rigorous, end-to-end theory of intelligent magnetic resonance imaging by unifying the physics of spin systems, the mathematics of inverse problems, and the modern machinery of artificial intelligence. Beginning from the quantum and semiclassical foundations of nuclear magnetic resonance, it builds systematically through Bloch dynamics, signal formation, Fourier encoding, image reconstruction, parameter estimation, uncertainty quantification, and the geometry of learned representations. Across the chapters, MRI is treated not simply as an imaging modality but as a layered computational-physical system in which measurements, models, and inference are inseparably linked. The book shows how classical tools such as Hilbert-space analysis, regularization theory, stochastic processes, and optimization evolve naturally into contemporary methods involving deep neural networks, graph models, generative priors, diffusion reconstruction, Bayesian inference, and operator learning. At the same time, the book argues that the future of MRI lies in systems that are not merely automated, but mathematically grounded, adaptive, and scientifically interpretable. It examines how AI can guide acquisition, reconstruction, multimodal fusion, motion correction, artifact suppression, quantitative imaging, and even agentic workflow orchestration across the entire MRI pipeline. Throughout, the emphasis remains on preserving physical consistency, clinical trustworthiness, and theoretical clarity while extending MRI into a new era of intelligent imaging. The result is a comprehensive research-level treatment of how magnetic resonance imaging is being transformed from a sequence of handcrafted procedures into a unified framework of physics-constrained, data-driven, and increasingly autonomous inference.