Based Magnetic Resonance Imaging Reconstruction

Deep learning-based magnetic resonance imaging (MRI) reconstruction aims to accelerate image acquisition and improve image quality by reconstructing high-resolution images from undersampled data. Current research focuses on integrating motion correction into reconstruction models, leveraging edge information and complex-valued networks for improved accuracy, and employing techniques like continual learning and contrastive learning to optimize performance for multiple downstream tasks and address fairness concerns. These advancements are significantly impacting clinical practice by enabling faster scans, reducing artifacts, and improving the diagnostic value of MRI, particularly in applications like intraoperative imaging.