Joint Learning of Cartesian under Sampling Andre Construction for Accelerated MRI

MRzero ‐ Automated discovery of MRI sequences using supervised learning

“…Most of this work deals with the problem of accelerated imaging, and attempts to learn optimal k-space undersampling patterns that minimize scan time, while still obeying scanner limitations and allowing for artifact-free reconstruction. [8][9][10][11][12] These approaches already link acquisition and reconstruction optimization, although the mentioned approaches only deal with the sampling pattern aspect of MR measurements and do not optimize for pulses, gradient forms, and timing. Novel approaches address the problem of general sequence learning.…”

Fast Data-Driven Learning of MRI Sampling Pattern for Large Scale Problems

“…However, joint optimization of the algorithm parameters and the SP may be a better approach to be investigated in the future. This is important for merging our approach with deep learning reconstruction, where learning sampling and reconstruction happen simultaneously, as recently in 53,54,55 . That appears to us likely to be the most suitable way for learned reconstruction approaches 17,19,20,83 , as well as for classical parametric inverse problems-based reconstructions 84,85,86 .…”

End-to-End Sequential Sampling and Reconstruction for MRI

“…More recently, deep learning has enabled a data-driven solution to the co-design problem, where the sampler and reconstructor can be jointly learned through end-to-end training. For example, [1,43,48] proposed co-design frameworks for 2D Cartesian κ-space sampling and [39,42] applied co-design to 2D radial κ-space sampling. 1 These methods have shown superior performance over previous baselines that combine an individually-optimized sampler and reconstructor pair [1,34,39,43,48].…”

“…For example, [1,43,48] proposed co-design frameworks for 2D Cartesian κ-space sampling and [39,42] applied co-design to 2D radial κ-space sampling. 1 These methods have shown superior performance over previous baselines that combine an individually-optimized sampler and reconstructor pair [1,34,39,43,48]. However, these methods do not take advantage of the sequential nature of data collection during an MRI scan, and only solve for a generic sampling pattern for an entire dataset.…”