Machine learning for medical image reconstruction : 5th international workshop, MLMIR 2022, held in conjunction with MICCAI 2022, Singapore, September 22, 2022 : proceedings

material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed. The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. The publisher, the authors, and the editors are safe to assume that the advice and information in this book are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors give a warranty, expressed or implied, with respect to the material contained herein or for any errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Preface 6 Organization 7 Contents 8 Deep Learning for Magnetic Resonance Imaging 10 Rethinking the Optimization Process for Self-supervised Model-Driven MRI Reconstruction 11 1 Introduction 12 2 Theory 13 2.1 Model Based Deep Learning Network 13 2.2 Self-supervised MRI Reconstruction 14 2.3 Derivation of K-Space Calibration 14 3 Experiments and Results 16 3.1 Dataset 16 3.2 Implementation Details 16 3.3 Results 16 4 Conclusion 19 References 19 NPB-REC: Non-parametric Assessment of Uncertainty in Deep-Learning-Based MRI Reconstruction from Undersampled Data 22 1 Introduction 22 2 Methods 24 2.1 MRI Reconstruction 24 2.2 Non-parametric Bayesian MRI Reconstruction 24 2.3 The Reconstruction Network 25 3 Experiments 26 3.1 Database 26 3.2 Experimental Setup 26 3.3 Results 27 4 Conclusions 29 References 30 Adversarial Robustness of MR Image Reconstruction Under Realistic Perturbations 32 1 Introduction 32 2 Methods 34 3 Experiments and Results 36 4 Conclusion 39 References 40 High-Fidelity MRI Reconstruction with the Densely Connected Network Cascade and Feature Residual Data Consistency Priors 42 1 Introduction 42 2 Method 43 2.1 Problem Formulation 43 2.2 Reconstruction Framework 44 2.3 Objective Function 46 3 Experiment 47 3.1 Comparison Results 47 3.2 Ablation Studies on Model Components 49 3.3 Ablation Studies on Bottleneck Design in DC Blocks 49 4 Conclusions and Discussion 50 References 50 Metal Artifact Correction MRI Using Multi-contrast Deep Neural Networks for Diagnosis of Degenerative Spinal Diseases 52 1 Introduction 52 2 Method 53 2.1 Data Preprocessing 53 2.2 Multi-contrast SEMAC Acceleration 55 2.3 Implementation Details 56 3 Experiment 56 3.1 Results of SEMAC Acceleration 56 3.2 Results of SEMAC/Phase-Encoding Acceleration 59 4 Discussion and Conclusion 59 References 60 Segmentation-Aware MRI Reconstruction 61 1 Introduction 61 2 Methods 62 2.1 Proposed Framework 62 2.2 Stabilization 62 2.3 Model Architectures 64 2.4 Implementation Details 64 3 Experimental Results 66 4 Conclusion 68 References 68 MRI Reconstruction with Conditional Adversarial Transformers 70 1 Introduction 70 2 Theory 71 2.1 Deep MRI Reconstruction 71 2.2 Conditional Adversarial Transformers 72 3 Methods 74 4 Results 75 5 Discussion 77 6 Conclusion 77 References 78 Deep Learning for General Image Reconstruction 80 A Noise-Level-Aware Framework for PET Image Denoising 81 1 Introduction 81 2 Noise-Level-Aware Framework 83 2.1 Quantification of Local Relative Noise Level 83 3 Experiments 85 3.1 Experimental Setup 85 3.2 Implementation Details 85 3.3 Results and Analysis 85 4 Conclusion 88 References 89 DuDoTrans: Dual-Domain Transformer for Sparse-View CT Reconstruction 90 1 Introduction and Motivation 90 2 Method 92 2.1 Network Architecture 92 3 Experimental Results 94 3.1 Ablation Study and Analysis 94 3.2 Sparse-View CT Reconstruction Analysis 96 4 Conclusion 98 References 98 Deep Denoising Network for X-Ray Fluoroscopic Image Sequences of Moving Objects 101 1 Introduction 101 2 Methods 103 2.1 Overall Architecture 103 2.2 Parallel Warping 105 2.3 Self-attention 105 2.4 Optimization of the Network 105 3 Experimental Results 106 3.1 Dataset Acquisition 106 3.2 Experimental Setup 106 3.3 Performance Comparison with State-of-the-Art Methods 107 3.4 Ablation Study 107 4 Discussion and Conclusion 108 References 109 PP-MPI: A Deep Plug-and-Play Prior for Magnetic Particle Imaging Reconstruction 111 1 Introduction 111 2 Background 113 2.1 MPI Signal Model 113 2.2 MPI Image Reconstruction 113 3 Methods 115 3.1 Plug-and-Play MPI Reconstruction (PP-MPI) 115 3.2 Analyses 117 4 Results 118 5 Discussion 119 References 119 Learning While Acquisition: Towards Active Learning Framework for Beamforming in Ultrasound Imaging 121 1 Introduction 121 2 Active Learning for US Beamforming 123 3 Results 126 4 Discussion and Conclusions 127 References 128 DPDudoNet: Deep-Prior Based Dual-Domain Network for Low-Dose Computed Tomography Reconstruction 129 1 Introduction 129 2 Method 130 2.1 The DPDudo Algorithm 130 2.2 The DPDudoNet 132 2.3 Interpretability of the DPDudoNet 132 2.4 Training Loss 133 3 Experimental Results 133 3.1 Clinical Data 133 3.2 Evaluation Metrics 133 3.3 Training Details 134 3.4 Performance Evaluation 134 3.5 Ablation Study 135 4 Conclusion 137 References 137 MTD-GAN: Multi-task Discriminator Based Generative Adversarial Networks for Low-Dose CT Denoising 139 1 Introduction 139 1.1 Deep Denoiser 140 1.2 Multi-task Learning 141 2 Methods 141 2.1 Multi-task Discriminator 141 2.2 Non-difference Suppression Loss and Consistency Loss 143 2.3 FFT-Generator 144 3 Experiments and Results 145 3.1 Experiments Settings 145 3.2 Comparison Results 146 4 Conclusion 147 References 148 Uncertainty-Informed Bayesian PET Image Reconstruction Using a Deep Image Prior 151 1 Introduction and Related Work 151 2 Methods 153 3 Data and Experiments 155 4 Results and Discussion 157 5 Conclusion 159 References 159 Author Index 162 This book constitutes the refereed proceedings of the 5th International Workshop on Machine Learning for Medical Reconstruction, MLMIR 2022, held in conjunction with MICCAI 2022, in September 2022, held in Singapore.The 15 papers presented were carefully reviewed and selected from 19 submissions. The papers are organized in the following topical sections: deep learning for magnetic resonance imaging and deep learning for general image reconstruction. This book constitutes the refereed proceedings of the 5th International Workshop on Machine Learning for Medical Reconstruction, MLMIR 2022, held in conjunction with MICCAI 2022, in September 2022, held in Singapore. The 15 papers presented were carefully reviewed and selected from 19 submissions. The papers are organized in the following topical sections: deep learning for magnetic resonance imaging and deep learning for general image reconstruction