Dynamic Magnetic Resonance

Dynamic magnetic resonance imaging (dMRI) aims to rapidly acquire high-resolution images of moving organs and tissues, overcoming limitations of traditional MRI's long scan times. Current research heavily utilizes deep learning, employing architectures like U-Nets, Swin Transformers, and unrolled iterative networks, often incorporating compressed sensing and various regularization techniques (e.g., low-rank, sparsity, smoothness) to improve reconstruction from undersampled k-space data. These advancements are crucial for improving diagnostic accuracy and patient comfort in applications like cardiac imaging and potentially expanding dMRI's use in real-time applications and other areas like esophageal health assessment. The field is also exploring self-supervised learning and the use of natural videos for training to address data scarcity issues.