Intro
Preface
Organization
Contents
Deep Learning for Magnetic Resonance Imaging
HyperRecon: Regularization-Agnostic CS-MRI Reconstruction with Hypernetworks
1 Introduction
2 Background
2.1 Amortized Optimization of CS-MRI
2.2 Hypernetworks
3 Proposed Method
3.1 Regularization-Agnostic Reconstruction Network
3.2 Training
4 Experiments
4.1 Hypernetwork Capacity and Hyperparameter Sampling
4.2 Range of Reconstructions
5 Conclusion
References
Efficient Image Registration Network for Non-Rigid Cardiac Motion Estimation
1 Introduction

2 Method
2.1 Network Architecture
2.2 Self-supervised Loss Function
2.3 Enhancement Mask (EM)
3 Experiments
4 Results
5 Discussion
6 Conclusion
References
26em plus .1em minus .1emEvaluation of the Robustness of Learned MR Image Reconstruction to Systematic Deviations Between Training and Test Data for the Models from the fastMRI Challenge*-6pt
1 Introduction
2 Methods
2.1 Image Perturbations
2.2 Description of 2019 fastMRI Approaches
3 Results
4 Discussion and Conclusion
References
Self-supervised Dynamic MRI Reconstruction
1 Introduction

2 Theory
2.1 Dynamic MRI Reconstruction
2.2 Self-supervised Learning
3 Methods
4 Experimental Results
5 Conclusion
References
A Simulation Pipeline to Generate Realistic Breast Images for Learning DCE-MRI Reconstruction
1 Introduction
2 Method
2.1 DCE-MRI Data Acquisition
2.2 Pharmacokinetics Model Analysis and Simulation
2.3 MR Acquisition Simulation
2.4 Testing with ML Reconstruction
3 Result
4 Discussion
5 Conclusion
References
Deep MRI Reconstruction with Generative Vision Transformers
1 Introduction
2 Theory

2.1 Deep Unsupervised MRI Reconstruction
2.2 Generative Vision Transformers
3 Methods
4 Results
5 Discussion
6 Conclusion
References
Distortion Removal and Deblurring of Single-Shot DWI MRI Scans
1 Introduction
2 Background
2.1 Distortion Removal Framework
2.2 EDSR Architecture
3 Distortion Removal and Deblurring of EPI-DWI
3.1 Data
3.2 Distortion Removal Using Structural Images
3.3 Pre-processing for Super-Resolution
3.4 Data Augmentation
3.5 Architectures Explored for EPI-DWI Deblurring
4 Experiments and Results

4.1 Computer Hardware Details
4.2 Training Details
4.3 Baselines
4.4 Evaluation Metrics
4.5 Results
5 Conclusion
References
One Network to Solve Them All: A Sequential Multi-task Joint Learning Network Framework for MR Imaging Pipeline
1 Introduction
2 Method
2.1 SampNet: The Sampling Pattern Learning Network
2.2 ReconNet: The Reconstruction Network
2.3 SegNet: The Segmentation Network
2.4 SemuNet: The Sequential Multi-task Joint Learning Network Framework
3 Experiments and Discussion
3.1 Experimental Details
3.2 Experiments Results