Information Processing in Medical Imaging: 27th International Conference, IPMI 2021, Virtual Event, June 28–June 30, 2021, Proceedings (Lecture Notes in Computer Science, 12729)

Description

This book constitutes the proceedings of the 27th International Conference on Information Processing in Medical Imaging, IPMI 2021, which was held online during June 28-30, 2021. The conference was originally planned to take place in Bornholm, Denmark, but changed to a virtual format due to the COVID-19 pandemic. The 59 full papers presented in this volume were carefully reviewed and selected from 200 submissions. They were organized in topical sections as follows: registration; causal models and interpretability; generative modelling; shape; brain connectivity; representation learning; segmentation; sequential modelling; learning with few or low quality labels; uncertainty quantification and generative modelling; and deep learning. Preface Organization Contents Registration HyperMorph: Amortized Hyperparameter Learning for Image Registration 1 Introduction 2 Related Work 3 Methods 3.1 HyperMorph 3.2 Hyperparameter Tuning 3.3 Implementation 4 Experiments 4.1 Experiment 1: HyperMorph Efficiency and Capacity 4.2 Experiment 2: Robustness to Initialization 4.3 Experiment 3: Hyperparameter-Tuning Utility 5 Conclusion References Deep Learning Based Geometric Registration for Medical Images: How Accurate Can We Get Without Visual Features? 1 Introduction 1.1 Related Work 1.2 Contribution 2 Methods 2.1 Loopy Belief Propagation for Regularised Registration of Keypoint Graphs 2.2 Geometric Feature Extraction with Graph Convolutional Neural Networks 2.3 Deep Learning Based End-to-End Geometric Registration Framework 2.4 Implementation Details: Keypoints, Visual Features and Integral Loss 3 Experiments and Results 4 Discussion and Conclusion References Diffeomorphic Registration with Density Changes for the Analysis of Imbalanced Shapes 1 Introduction 2 Diffeomorphic Registration of Geometric Measures 2.1 Diffeomorphisms and Registration 2.2 Geometric Measure Representation of Shapes 3 Diffeomorphic Registration with Density Changes 3.1 An Augmented Optimal Control Problem 3.2 Numerical Implementation 3.3 Local Density Changes 4 Results 5 Conclusion References Causal Models and Interpretability Estimation of Causal Effects in the Presence of Unobserved Confounding in the Alzheimer's Continuum 1 Introduction 2 Methods 2.1 The Causal Question and Its Associated Graph 2.2 Identifiability in the Presence of an Unobserved Confounder 2.3 Estimating a Substitute Confounder 2.4 Identifiability in the Presence of a Substitute Confounder 2.5 The Outcome Model 3 Experiments 4 Conclusion References Multiple-Shooting Adjoint Method for Whole-Brain Dynamic Causal Modeling 1 Introduction 2 Methods 2.1 Notations and Formulation of Problem 2.2 Multiple-Shooting Method 2.3 Adjoint State Method 2.4 Multiple-Shooting Adjoint (MSA) Method 2.5 Dynamic Causal Modeling 3 Experiments 3.1 Validation on Toy Examples 3.2 Application to Whole-Brain Dynamic Causal Modeling with fMRI 4 Conclusion References Going Beyond Saliency Maps: Training Deep Models to Interpret Deep Models 1 Introduction 2 Related Work 3 Methods 3.1 Cycle-Consistent Image Simulation 3.2 Coupling Simulators via Conditional Convolution 3.3 Learning Warping Fields 4 Experiments 4.1 Synthetic Experiments 4.2 Visualizing the Effect of Alzheimer's Disease 4.3 Visualizing the Effect of Alcohol Dependence 5 Conclusion References Generative Modelling Enabling Data Diversity: Efficient Automatic Augmentation via Regularized Adversarial Training 1 Introduction 2 Related Work 3 Method 3.1 Preliminaries 3.2 Regularized Adversarial Data Augmentation 3.3 Data Augmentation Space and Models 4 Experiments 4.1 Experiments Setup 4.2 Skin Lesion Diagnosis Result 4.3 Organ-at-Risk Segmentation Result 5 Conclusion References Blind Stain Separation Using Model-Aware Generative Learning and Its Applications on Fluorescence Microscopy Images 1 Introduction 2 Methodology 3 Fluorescence Unmixing 4 Experimental Evaluation 4.1 Experimental Setup 4.2 Results and Discussions 5 Conclusions References MR Slice Profile Estimation by Learning to Match Internal Patch Distributions 1 Introduction 2 Methods 2.1 Slice Profile 2.2 Slice Profile and Internal Patch Distributions 2.3 Slice Profile and GAN 2.4 Regularization Functions and Other Details 3 Experiments and Results 3.1 Simulations from Isotropic Images 3.2 Incorporating Slice Profile Estimation into SMORE 3.3 Measuring Through-Plane Resolution After Applying SMORE 4 Discussion and Conclusions References Shape Partial Matching in the Space of Varifolds 1 Introduction 2 Partial Matching 2.1 The Varifold Framework for Shape Matching 2.2 Definition of the Partial Matching Dissimilarity 2.3 Normalized Partial Matching Dissimilarity 2.4 Use in the LDDMM Setting 3 Experiments 4 Conclusion References Nested Grassmanns for Dimensionality Reduction with Applications to Shape Analysis 1 Introduction 2 Nested Grassmannians 2.1 The Riemannian Geometry of Grassmann Manifolds 2.2 Embedding of Gr(p, m) in Gr(p, n) 2.3 Unsupervised Dimensionality Reduction 2.4 Supervised Dimensionality Reduction 2.5 Choice of the Distance d 2.6 Analysis of Principal Nested Grassmanns 3 Experiments 3.1 Synthetic Data 3.2 Application to Planar Shape Analysis 4 Conclusion References Hierarchical Morphology-Guided Tooth Instance Segmentation from CBCT Images 1 Introduction 2 Methods 2.1 Tooth Centroid and Skeleton Extraction Network 2.2 Multi-task Learning for Tooth Segmentation 2.3 Implementation Details 3 Experimental Results 3.1 Dataset and Evaluation Metrics 3.2 Evaluation and Comparison 3.3 Comparison with the State-of-the-Art Methods 4 Conclusion References Cortical Morphometry Analysis Based on Worst Transportation Theory 1 Introduction 2 Theoretic Results 2.1 Optimal Transportation Map 2.2 Worst Transportation Map 2.3 Geometric Variational Method 3 Computational Algorithms 3.1 Basic Concepts from Computational Geometry 3.2 Algorithms Based on Computational Geometry 4 Experiments 5 Conclusion References Geodesic B-score for Improved Assessment of Knee Osteoarthritis 1 Introduction 2 Background 2.1 Shape Space 2.2 Geometric Statistics 3 Geodesic B-score 3.1 Generalization 3.2 Sex-Specific Reference 3.3 Algorithmic Treatment 4 Results and Discussion 4.1 Data Description 4.2 Efficiency of Projection Algorithm 4.3 Predictive Validity 5 Conclusion and Future Work References Brain Connectivity Cytoarchitecture Measurements in Brain Gray Matter Using Likelihood-Free Inference 1 Introduction 2 Methods 2.1 Modeling the Brain Gray Matter with a 3-Compartment Model 2.2 An Invertible 3-Compartment Model: dMRI Summary Statistics 2.3 Solving the Inverse Problem via Likelihood Free Inference 3 Results and Discussion 3.1 Simulations 3.2 HCP MGH Results 4 Conclusion References Non-isomorphic Inter-modality Graph Alignment and Synthesis for Holistic Brain Mapping 1 Introduction 2 Methodology 3 Experimental Results and Discussion 4 Conclusion References Knowledge Transfer for Few-Shot Segmentation of Novel White Matter Tracts 1 Introduction 2 Methods 2.1 Problem Formulation and Classic Fine-Tuning 2.2 Knowledge Transfer for Few-Shot Segmentation of Novel WM Tracts 2.3 A Better Implementation with Warmup 2.4 Implementation Details 3 Results 3.1 Data Description and Experimental Settings 3.2 Evaluation of Segmentation Accuracy 3.3 Impact of the Number of Annotated Training Scans 3.4 Evaluation of Volume Difference 4 Discussion 5 Conclusion References Discovering Spreading Pathways of Neuropathological Events in Alzheimer’s Disease Using Harmonic Wavelets 1 Introduction 2 Methods 2.1 Manifold Harmonics 2.2 Construction of Region-Adaptive Harmonic Wavelets 3 Experiments 3.1 Evaluate the Representation Power on Harmonic Wavelets 3.2 Evaluate the Statistic Power of Harmonic Wavelet Fingerprint 4 Conclusions References A Multi-scale Spatial and Temporal Attention Network on Dynamic Connectivity to Localize the Eloquent Cortex in Brain Tumor Patients 1 Introduction 2 A Multi-scale Spatial and Temporal Attention Network to Localize the Eloquent Cortex 2.1 Input Dynamic Connectivity Matrices 2.2 Multi-scale Spatial Attention on Convolutional Features 2.3 Temporal Attention Model and Multi-task Learning 3 Experimental Results 3.1 Dataset and Preprocessing 3.2 Localization Results 3.3 Feature Analysis 4 Conclusion References Learning Multi-resolution Graph Edge Embedding for Discovering Brain Network Dysfunction in Neurological Disorders 1 Introduction 1.1 Related Work 2 Proposed Method 2.1 Multi-resolution Graph Edge Transform 2.2 Efficient Graph Matrix Transform 2.3 Network Architecture 2.4 Training MENET 3 Experiments 3.1 Datasets 3.2 Experimental Settings 3.3 Structural Brain Connectivity Analysis on ADNI 3.4 Functional Brain Connectivity Analysis on ADHD 3.5 Discussions on Convergence of Scales 4 Conclusion References Equivariant Spherical Deconvolution: Learning Sparse Orientation Distribution Functions from Spherical Data 1 Introduction 2 Methods 2.1 Background and Preliminaries 2.2 Equivariant Spherical Deconvolution 3 Experiments 3.1 Noisy Synthetic Benchmark 3.2 The Tractometer Benchmark 3.3 Real-World Multi-shell Human Dataset 4 Discussion References Geodesic Tubes for Uncertainty Quantification in Diffusion MRI 1 Introduction 2 Theory 3 Experiments 4 Discussion References Structural Connectome Atlas Construction in the Space of Riemannian Metrics 1 Introduction 2 Structural Connectomes as Riemannian Metrics 3 The Geometry of the Manifold of All Metrics 3.1 The Induced Distance Function on the Diffeomorphism Group 4 Computational Anatomy of the Human Connectome 4.1 Estimating the Atlas for a Population of Connectomes 4.2 Implementation Details 5 Results 6 Conclusions References A Higher Order Manifold-Valued Convolutional Neural Network with Applications to Diffusion MRI Processing 1 Introduction 2 Preliminary 3 Manifold-Valued Volterra Series and Convolution 3.1 Manifold-Valued Volterra Series 3.2 Manifold-Valued Deep Network Based on MVVS/MVC 3.3 The Cases of Sn and SPD(n) 4 Experiments 4.1 Parkinson's Disease vs. Controls Classification 4.2 fODF Reconstruction 5 Conclusion References Representation Learning Representation Disentanglement for Multi-modal Brain MRI Analysis 1 Introduction 2 Related Works 3 Proposed Method 3.1 Representation Disentanglement by Image-to-image Translation 3.2 Fusing Disentangled Representations for Downstream Tasks 4 Experiments 4.1 Datasets 4.2 Experimental Settings 4.3 Evaluation on Disentangled Representation 4.4 Evaluation on Downstream Tasks 5 Conclusion References Variational Knowledge Distillation for Disease Classification in Chest X-Rays 1 Introduction 2 Related Work 3 Methodology 3.1 Preliminaries 3.2 Disease Classification by Conditional Variational Inference 3.3 Knowledge Distillation from EHRs 3.4 Empirical Objective Function 3.5 Implementation with Neural Networks 4 Experiments 4.1 Datasets 4.2 Experimental Settings 4.3 Results 5 Conclusion References Information-Based Disentangled Representation Learning for Unsupervised MR Harmonization 1 Introduction 2 Method 2.1 The Disentangling Framework 2.2 Creating a Consistent Anatomical Space 2.3 Learning from an Information Bottleneck 2.4 Domain Adaptation 3 Experiments and Results 3.1 Datasets and Preprocessing 3.2 Qualitative and Quantitative Evaluation 3.3 Domain Adaptation 4 Discussion and Conclusion References A3DSegNet: Anatomy-Aware Artifact Disentanglement and Segmentation Network for Unpaired Segmentation, Artifact Reduction, and Modality Translation 1 Introduction 2 Methodology 2.1 Network Architecture 2.2 Network Learning and Loss Functions 2.3 Anatomy-Aware Modality Translation 2.4 3D Segmentation 3 Experiments 3.1 Dataset and Experiment Setup 3.2 Ablation Study 3.3 Comparison with State-of-the-Art 4 Conclusions References Unsupervised Learning of Local Discriminative Representation for Medical Images 1 Introduction 2 Related Work 3 Methods 3.1 Local Discrimination Learning 3.2 Prior-Guided Anatomical Structure Clustering 4 Experiments and Discussion 4.1 Network Architectures and Initialization 4.2 Experiments for Learning Local Discrimination 4.3 Experiments for Clustering Structures Based on Prior Knowledge 5 Conclusion References TopoTxR: A Topological Biomarker for Predicting Treatment Response in Breast Cancer 1 Introduction 1.1 Related Work 2 Methodology 2.1 Background: Persistent Homology 2.2 Persistence Cycles and Their Computation 2.3 Topological-Cycle-Driven 3D CNN 3 Experimental Results 3.1 Discussion and TopoTxR Feature Interpretation 4 Conclusion References Segmentation Segmenting Two-Dimensional Structures with Strided Tensor Networks 1 Introduction 2 Methods 2.1 Overview 2.2 Tensor Notation 2.3 Image Segmentation Using Linear Models 2.4 Strided Tensor Networks 2.5 Optimisation 3 Data and Experiments 3.1 Data 3.2 Experiments 3.3 Results 4 Discussion and Conclusions References Distributional Gaussian Process Layers for Outlier Detection in Image Segmentation 1 Introduction 1.1 Related Work 1.2 Contributions 2 Hierarchical GP with Wasserstein-2 Kernels 3 Convolutionally Warped DistGP and Activation Function 4 Imposing Lipschitz Conditions in Convolutionally Warped DistGP 4.1 Proving Lipschitz Bounds in a DistGP Layer 5 DistGP-Based Segmentation Network and OOD Detection 6 Evaluation on Brain MRI 6.1 Data and Pre-processing 6.2 Brain Tissue Segmentation on Normal MRI Scans 6.3 Outlier Detection in MRI Scans with Tumors 7 Discussion References Deep Label Fusion: A 3D End-To-End Hybrid Multi-atlas Segmentation and Deep Learning Pipeline 1 Introduction 2 Materials 3 Method 3.1 Deep Label Fusion 3.2 Alternative Methods for Comparisons 4 Experiments and Results 5 Conclusion References Feature Library: A Benchmark for Cervical Lesion Segmentation 1 Introduction 2 Related Works 2.1 Automated Cervical Lesion Segmentation 2.2 Colposcopic Image Datasets 3 CINEMA Dataset 4 Benchmark Method 5 Experiments 5.1 Ablation Study 5.2 Comparison with Baselines 6 Conclusion References Generalized Organ Segmentation by Imitating One-Shot Reasoning Using Anatomical Correlation 1 Introduction 2 Related Work 3 Proposed Method 3.1 One-Shot Reasoning Using OrganNet 3.2 Pyramid Reasoning Modules 4 Experiments and Results 4.1 Dataset and Evaluation Metric 4.2 Implementation Details 4.3 Comparison with One-Shot Segmentation Methods: Better Performance 4.4 Comparison with Supervised 3D U-Nets: Less Labeling Cost 4.5 Ablation Study on Network Design 4.6 Visual Analysis 4.7 Generalization to Non-organ Segmentation 5 Conclusions and Future Work References EnMcGAN: Adversarial Ensemble Learning for 3D Complete Renal Structures Segmentation 1 Introduction 2 Methodology 2.1 Multi-windowing Committee for Multiple Fine-Grained Representation 2.2 Multi-condition GAN for Shape Constraints 2.3 Adversarial Weighted Ensemble for Personalized Fine Fusion 3 Materials and Configurations 4 Results and Analysis 4.1 Comparative Study Shows Superiority 4.2 Ablation Study Shows Improvements of the Innovations 4.3 Framework Analysis 5 Conclusion References Segmentation with Multiple Acceptable Annotations: A Case Study of Myocardial Segmentation in Contrast Echocardiography 1 Introduction 2 Related Work 3 Method 4 Experiment 4.1 Dataset 4.2 Training DNN with Extended Dice Loss 4.3 Extended Dice as a Superior Evaluation Metric 5 Conclusion References A New Bidirectional Unsupervised Domain Adaptation Segmentation Framework 1 Introduction 2 Method 2.1 Problem Definition 2.2 DRPL Framework with Domain-Aware Pattern Encoder 2.3 Loss Functions for DRPL-based BiUDA Framework 3 Experiments 4 Discussion 5 Conclusion References 3D Nucleus Instance Segmentation for Whole-Brain Microscopy Images 1 Introduction 2 Problem Formulation 3 Proposed Model 3.1 Encoder 3.2 Reconstruction Branch 3.3 Identity Propagation Branch 4 Experiments 4.1 Dataset 4.2 Experiment Setup 4.3 Results 5 Conclusion References Teach Me to Segment with Mixed Supervision: Confident Students Become Masters 1 Introduction 2 Related Work 3 Methodology 3.1 Multi-branch Architecture 3.2 Supervised Learning 3.3 Not So-Supervised Branch 3.4 Distilling Strong Knowledge 3.5 Shannon-Entropy Minimization 3.6 Link Between Entropy and Pseudo-mask Supervision 3.7 Joint Objective 4 Experimental Setting 4.1 Results References Sequential Modelling Future Frame Prediction for Robot-Assisted Surgery 1 Introduction 2 Method 2.1 Problem Formulation 2.2 Decomposed Video Encoding Network 2.3 Ternary Latent Variable 2.4 Learning Process 3 Experimental Results 3.1 Dataset and Evaluation Metrics 3.2 Implementation Details 3.3 Results 4 Conclusion and Future Work References Velocity-To-Pressure (V2P) - Net: Inferring Relative Pressures from Time-Varying 3D Fluid Flow Velocities 1 Introduction 2 Methodology 2.1 Background 2.2 V2P-Net: 3 Experiments 4 Results and Discussion: 5 Conclusion References Lighting Enhancement Aids Reconstruction of Colonoscopic Surfaces 1 Introduction 2 Background: SLAM and Colonoscopic Reconstruction 2.1 SLAM Mechanism 2.2 The Lighting Problem in Colonoscopic Surface Reconstruction 3 Method 3.1 Adaptive Gamma Correction 3.2 RNN Network 3.3 Training Strategy 4 Experiments 4.1 Implementation Details 4.2 Visual Effect 4.3 Application in Colonoscopic Surface Reconstruction 5 Discussion and Conclusion References Mixture Modeling for Identifying Subtypes in Disease Course Mapping 1 Introduction 2 Related Work 3 Method 3.1 Disease Course Mapping Model 3.2 Mixture of Disease Course Mapping Models 3.3 Tempered Scheme 3.4 Initialization Method 4 Results and Discussion 4.1 Simulated Data 4.2 Applications on Alzheimer's Disease Data References Learning Transition Times in Event Sequences: The Temporal Event-Based Model of Disease Progression 1 Introduction 2 Theory 2.1 Temporal Event-Based Model 2.2 Inference 2.3 Staging 3 Experiments and Results 3.1 Alzheimer's Disease Data 3.2 Model Training 3.3 TEBM Parameters 3.4 Alzheimer's Disease Timeline 3.5 Individual Trajectories 3.6 Prediction of Progression Rate 3.7 Comparative Model Performance 3.8 Performance with Missing Data 4 Discussion References Learning with Few or Low Quality Labels Knowledge Distillation with Adaptive Asymmetric Label Sharpening for Semi-supervised Fracture Detection in Chest X-Rays 1 Introduction 2 Method 2.1 Knowledge Distillation Learning 2.2 Adaptive Asymmetric Label Sharpening 2.3 Implementation Details 3 Experiments 3.1 Experimental Settings 3.2 Comparison with Baseline Methods 3.3 Ablation Study 4 Conclusion References Semi-Supervised Screening of COVID-19 from Positive and Unlabeled Data with Constraint Non-Negative Risk Estimator 1 Introduction 2 Related Work 2.1 Automated COVID-19 Screening 2.2 Positive Unlabeled Learning 3 Methodology 3.1 Learning Set-Up 3.2 Unbiased PU Learning 3.3 Constraint Non-Negative PU Learning 3.4 Theoretical Analyses 4 Experiments 4.1 Setup 4.2 Results 4.3 Insight Analyses 5 Conclusion References Deep MCEM for Weakly-Supervised Learning to Jointly Segment and Recognize Objects Using Very Few Expert Segmentations 1 Introduction and Related Work 2 Methods 2.1 DNN-Based Variational Model for Object Segmentation 2.2 DNN-Based Statistical Model to Recognize a Segmented Object 2.3 MCEM for Weakly-Supervised Segmenter-Recognizer Learning 2.4 Efficient MH Sampling of Missing Segmentations and Encodings 2.5 Deep-MCEM Based Inference Strategy for Test Images 3 Results and Discussion 4 Conclusion References Weakly Supervised Deep Learning for Aortic Valve Finite Element Mesh Generation from 3D CT Images 1 Introduction 2 Methods 2.1 Possible Problem Formulation: Meshing from Segmentation 2.2 Proposed Problem Formulation: Mesh Template Matching 2.3 Deformation Field 3 Experiments and Results 3.1 Data Acquisition and Preprocessing 3.2 Implementation Details 3.3 Evaluation Metrics 3.4 Comparison with an Image Intensity Gradient-Based Approach 3.5 Comparison with Other Deformation Strategies 3.6 FEA Results 3.7 Limitations and Future Works 4 Conclusion References Continual Active Learning for Efficient Adaptation of Machine Learning Models to Changing Image Acquisition 1 Introduction 2 Method 2.1 CASA Training Scheme 2.2 Pseudo-domain Module 2.3 Task Module 2.4 Training Memory 2.5 Outlier Memory and Pseudo-domain Identification 3 Experiments and Results 3.1 Data Set 3.2 Methods Compared in the Evaluation 3.3 Experimental Setup 3.4 Model Accuracy Across Domains 3.5 Evaluation of the Memory and Pseudo-Domains 4 Conclusion References Multimodal Self-supervised Learning for Medical Image Analysis 1 Introduction 2 Related Work 3 Method 3.1 Multimodal Puzzle Construction 3.2 Puzzle-Solving with Sinkhorn Networks 3.3 Cross-modal Generation 4 Experimental Results 4.1 Datasets 4.2 Transfer Learning ResultsWe Evaluate on realistic data in this section, using a 5-fold cross validation approach. 4.3 Low-Shot Learning Results 4.4 Cross-modal Generation Results 5 Conclusion and Future Work References Uncertainty Quantification and Generative Modelling Spatially Varying Label Smoothing: Capturing Uncertainty from Expert Annotations 1 Introduction 2 Spatially Varying Label Smoothing 2.1 Multi-rater SVLS 3 Experiments 3.1 Datasets 3.2 Implementation Details and Baselines 4 Results 4.1 Evaluation Metrics 4.2 Multi-class Image Segmentation 4.3 Multiple Expert Annotations 5 Discussion References Quantile Regression for Uncertainty Estimation in VAEs with Applications to Brain Lesion Detection 1 Introduction 2 Background 2.1 Variance Shrinkage Problem in Variational Autoencoders 2.2 Conditional Quantile Regression 3 Proposed Approach: Uncertainty Estimation for Autoencoders with Quantile Regression (QR-VAE) 4 Experiments and Results 4.1 Simulations 4.2 Unsupervised Lesion Detection 5 Conclusion References A Probabilistic Framework for Modeling the Variability Across Federated Datasets 1 Introduction 2 State of the Art 3 Federated Multi-views PPCA 3.1 Problem Setup 3.2 Modeling Assumptions 3.3 Proposed Framework 4 Applications 4.1 Materials 4.2 Benchmark 4.3 Results 5 Conclusions References Is Segmentation Uncertainty Useful? 1 Introduction 2 Modelling Segmentation Uncertainty 3 Probabilistic Segmentation Networks 4 Experiments 4.1 Data 4.2 Model Tuning and Training 4.3 Uncertainty Estimation 4.4 Sampling Segmentation Masks 4.5 Uncertainty Estimates for Active Learning 5 Discussion and Conclusion References Deep Learning Principled Ultrasound Data Augmentation for Classification of Standard Planes 1 Introduction 2 Methods 3 Experiments and Results 4 Conclusion References Adversarial Regression Learning for Bone Age Estimation 1 Introduction 2 Related Work 3 Method 3.1 Motivation 3.2 Problem 3.3 Adversarial Regression Learning 3.4 Overall Objective 4 Experiments 4.1 Datasets 4.2 Implementation Details 4.3 Evaluation 5 Discussion 6 Conclusion References Learning Image Quality Assessment by Reinforcing Task Amenable Data Selection 1 Introduction 2 Method 2.1 Image Quality Assessment by Task Amenability 2.2 The Reinforcement Learning Algorithm 2.3 Image Quality Assessment with Reinforcement Learning 3 Experiment 4 Result 5 Discussion and Conclusion References Collaborative Multi-agent Reinforcement Learning for Landmark Localization Using Continuous Action Space 1 Introduction 2 Method 3 Experimental Setup 4 Results 5 Discussion 6 Conclusion References Author Index