Preface
Organization
Contents – Part II
Optical and Histology Applications: Optical Imaging Applications
Instance Segmentation and Tracking with Cosine Embeddings and Recurrent Hourglass Networks
1 Introduction
2 Instance Segmentation and Tracking
2.1 Recurrent Stacked Hourglass Network
2.2 Cosine Embedding Loss
2.3 Clustering of Embeddings
3 Experimental Setup and Results
4 Discussion and Conclusion
References
Skin Lesion Classification in Dermoscopy Images Using Synergic Deep Learning
1 Introduction
2 Datasets
3 Method
3.1 Input Layer
3.2 Dual DCNN Components
3.3 Synergic Network
3.4 Training and Testing
4 Results
5 Discussion
6 Conclusion
References
SLSDeep: Skin Lesion Segmentation Based on Dilated Residual and Pyramid Pooling Networks
1 Introduction
2 Proposed Model
2.1 Network Architecture
2.2 Loss Function
3 Experimental Setup and Evaluation
4 Conclusions
References
-Hemolysis Detection on Cultured Blood Agar Plates by Convolutional Neural Networks
1 Introduction
1.1 Problem Definition
1.2 Related Work and Contribution
2 Throat Swab Culture Dataset
3 -Hemolysis Detection Method
3.1 Patch Extraction
3.2 Patch Classification
4 Results and Discussion
5 Conclusion
References
A Pixel-Wise Distance Regression Approach for Joint Retinal Optical Disc and Fovea Detection
1 Introduction
2 Joint Optic Disc and Fovea Detection Methodology
2.1 Casting the Problem into a Pixel-Wise Regression Task
2.2 A Fully-Convolutional Deep Neural Network for Distance Regression
2.3 Optic Disc and the Fovea Assignment
3 Experimental Evaluation
3.1 Data
3.2 Evaluation Approach
3.3 Quantitative Evaluation
4 Conclusions and Future Work
References
Deep Random Walk for Drusen Segmentation from Fundus Images
1 Introduction
2 Deep Random Walk Networks
2.1 Deep Feature Extraction Module
2.2 Affinity Learning Module
2.3 Random Walk Module
3 Implementation
4 Experiment
4.1 Dataset
4.2 Evaluation and Result
5 Conclusion
References
Retinal Artery and Vein Classification via Dominant Sets Clustering-Based Vascular Topology Estimation
1 Introduction
2 Method
2.1 Graph Generation
2.2 Dominant Sets Clustering-Based Topology Estimation
2.3 Artery/Vein Classification
3 Experimental Results
3.1 Topology Estimation
3.2 A/V Classification
4 Conclusions
References
Towards a Glaucoma Risk Index Based on Simulated Hemodynamics from Fundus Images
1 Introduction
2 Materials and Methods
2.1 Preprocessing
2.2 Simulation of the Retinal Hemodynamics
2.3 Bag of Hemodynamic Features (BoHF)
3 Results
4 Discussion
References
A Framework for Identifying Diabetic Retinopathy Based on Anti-noise Detection and Attention-Based Fusion
1 Introduction
2 Method
2.1 Center-Sample Detector
2.2 Attention Fusion Network
3 Experiments
3.1 Evaluating Center-Sample Detector
3.2 Evaluating the Attention Fusion Network
4 Conclusions
References
Deep Supervision with Additional Labels for Retinal Vessel Segmentation Task
1 Introduction
2 Proposed Method
2.1 U-Net
2.2 Additional Label
2.3 Deep Supervision
3 Experiments
3.1 Datasets
3.2 Results
3.3 Comparison
4 Conclusion
References
A Multi-task Network to Detect Junctions in Retinal Vasculature
1 Introduction
2 Method
2.1 Learning Junction Patterns with Multi-task Network
2.2 Initial Junction Search
2.3 Refined Junction Search
3 Experimental Setup and Material
4 Results
5 Conclusion
References
A Multitask Learning Architecture for Simultaneous Segmentation of Bright and Red Lesions in Fundus Images
1 Introduction
2 Methods
2.1 Multitask Architecture
2.2 Training
3 Experiments
4 Results and Discussion
5 Conclusion
References
Uniqueness-Driven Saliency Analysis for Automated Lesion Detection with Applications to Retinal Diseases
1 Introduction
2 Method
3 Experimental Evaluation
3.1 Dark Lesion Detection
3.2 Bright Lesion Detection
4 Conclusions
References
Multiscale Network Followed Network Model for Retinal Vessel Segmentation
1 Introduction
2 Datasets
3 Method
3.1 Images Pre-processing and Patch Extraction
3.2 Training Two NFN Models
3.3 Testing the MS-NFN Model
4 Results
5 Conclusions
References
Optical and Histology Applications: Histology Applications
Predicting Cancer with a Recurrent Visual Attention Model for Histopathology Images
1 Introduction
2 Method
3 Experiments
4 Conclusion
References
A Deep Model with Shape-Preserving Loss for Gland Instance Segmentation
1 Introduction
2 Method
2.1 Shape-Preserving Loss
2.2 Deep Convolutional Neural Network
3 Evaluation and Discussion
4 Conclusion
References
Model-Based Refinement of Nonlinear Registrations in 3D Histology Reconstruction
1 Introduction
2 Methods
2.1 Probabilistic Framework
2.2 Inference: Proposed Method
3 Experiments and Results
3.1 Data
3.2 Experiments on Synthetic Dataset
3.3 Experiments on Allen Dataset
4 Discussion and Conclusion
References
Invasive Cancer Detection Utilizing Compressed Convolutional Neural Network and Transfer Learning
1 Introduction
2 Methodology
2.1 Overview
2.2 Small Capacity Network
2.3 Transfer Learning from Large Capacity Network
2.4 Efficient Inference
3 Experiment and Discussion
4 Conclusion
References
Which Way Round? A Study on the Performance of Stain-Translation for Segmenting Arbitrarily Dyed Histological Images
1 Motivation
2 Methods
2.1 Stain-Translation Model and Sampling Strategies
2.2 Segmentation Model and Evaluation Details
3 Results
4 Discussion
References
Graph CNN for Survival Analysis on Whole Slide Pathological Images
1 Introduction
2 Methodology
3 Experiment
3.1 Dataset
3.2 State-of-the-Art Methods
3.3 Result and Discussion
4 Conclusion
References
Fully Automated Blind Color Deconvolution of Histopathological Images
1 Introduction
2 Problem Formulation
3 Bayesian Modelling and Inference
4 Experiments
5 Conclusions
References
Improving Whole Slide Segmentation Through Visual Context - A Systematic Study
1 Introduction
2 Related Work
3 Methods
4 Results and Discussion
5 Conclusions
References
Adversarial Domain Adaptation for Classification of Prostate Histopathology Whole-Slide Images
1 Introduction
2 Method
3 Experimental Validation and Results
4 Conclusion
References
Rotation Equivariant CNNs for Digital Pathology
1 Introduction
2 Methods
2.1 Background
2.2 G-CNN DenseNet Architecture
3 Experimental Results
3.1 Datasets and Evaluation
3.2 Model Reliability
3.3 P4M-DenseNet Performance
4 Conclusion
References
A Probabilistic Model Combining Deep Learning and Multi-atlas Segmentation for Semi-automated Labelling of Histology
1 Introduction
2 Methods
2.1 Probabilistic Model
2.2 Inference: Proposed Method
2.3 Model Instantiation
3 Experiments and Results
3.1 Data
3.2 Experimental Setup
3.3 Results
4 Discussion and Conclusion
References
BESNet: Boundary-Enhanced Segmentation of Cells in Histopathological Images
1 Introduction
2 Method
2.1 Boundary-Enhanced Segmentation Network (BESNet)
2.2 Boundary-Enhanced Cross-Entropy (BECE) Loss
2.3 Training and Testing
3 Experiments
4 Results and Discussions
5 Conclusions
References
Panoptic Segmentation with an End-to-End Cell R-CNN for Pathology Image Analysis
1 Introduction
2 Methods
2.1 Semantic Segmentation Branch and Feature Map Branch
2.2 Region Proposal Network and Instance Branch
3 Experiment
4 Conclusions
References
Integration of Spatial Distribution in Imaging-Genetics
1 Introduction
2 Method
2.1 Extraction of Image Features
2.2 Computation of Spatial Descriptor
2.3 Canonical Correlation Analysis
3 Experiments and Results
3.1 Ripley's K-Function on Real Data
3.2 CCA with Image and Spatial Features
3.3 Sparse CCA with Image and Spatial Features
4 Discussion and Conclusions
References
Multiple Instance Learning for Heterogeneous Images: Training a CNN for Histopathology
1 Introduction
2 Background
3 Multiple Instance Learning with a CNN
4 Multiple Instance Aggregation
5 Training with Multiple Instance Augmentation
6 Experiments
7 Discussion
References
Optical and Histology Applications: Microscopy Applications
Cell Detection with Star-Convex Polygons
1 Introduction
2 Method
2.1 Implementation
3 Experiments
3.1 Datasets
3.2 Evaluation Metric
3.3 Compared Methods
3.4 Results
4 Discussion
References
Deep Convolutional Gaussian Mixture Model for Stain-Color Normalization of Histopathological Images
1 Introduction
2 Methods
3 Experimental Results
4 Conclusions
References
Learning to Segment 3D Linear Structures Using Only 2D Annotations
1 Introduction
2 Related Work
3 Method
3.1 From 3D to 2D Annotations
3.2 Visual Hull for Training on Cropped Volumes
3.3 Implementation
4 Experimental Evaluation
4.1 Data and Annotations
4.2 User Study
4.3 3D vs 2D Annotations
5 Conclusion
References
A Multiresolution Convolutional Neural Network with Partial Label Training for Annotating Reflectance Confocal Microscopy Images of Skin
1 Introduction
2 Related Work
3 Proposed Model
3.1 Architecture
3.2 Loss Function
4 Dataset and Experiments
References
Weakly-Supervised Learning-Based Feature Localization for Confocal Laser Endomicroscopy Glioma Images
Abstract
1 Introduction
2 Methods
2.1 New Design of Class Activation Map (CAM)
2.2 Lateral Inhibition and Collateral Integration of Localizer Neurons
3 Experimental Setup and Results
4 Conclusions
Acknowledgement
References
Synaptic Partner Prediction from Point Annotations in Insect Brains
1 Introduction
2 Method
2.1 Directed edges for synaptic partner representation
2.2 Edge classification
2.3 Synaptic partner extraction
3 Results
4 Discussion
References
Synaptic Cleft Segmentation in Non-isotropic Volume Electron Microscopy of the Complete Drosophila Brain
1 Introduction
2 Related Work
3 Methods
3.1 Training Setup
3.2 Experiments
3.3 Synaptic Cleft Prediction on the Complete Drosophila Brain
4 Conclusion
References
Weakly Supervised Representation Learning for Endomicroscopy Image Analysis
1 Introduction
2 Methodology
2.1 Frame-Based Feature Representation
2.2 Local Label Classification
2.3 Global Label Classification
2.4 Semantic Exclusivity Loss
2.5 Final Objective and Alternative Learning
3 Experiments
4 Conclusion
References
DeepHCS: Bright-Field to Fluorescence Microscopy Image Conversion Using Deep Learning for Label-Free High-Content Screening
1 Introduction
2 Method
2.1 Data
2.2 Proposed Method: DeepHCS
3 Results
4 Conclusion
References
Optical and Histology Applications: Optical Coherence Tomography and Other Optical Imaging Applications
A Cascaded Refinement GAN for Phase Contrast Microscopy Image Super Resolution
1 Introduction
2 Related Work and Our Proposal
3 Preliminaries
3.1 Generative Adversarial Networks
3.2 Optics-Related Data Enhancement
4 A Cascaded Refinement GAN for Super Resolution
4.1 Network Architecture
4.2 Loss Function
4.3 Implementation and Training Details
5 Experimental Results
6 Conclusion
References
Multi-context Deep Network for Angle-Closure Glaucoma Screening in Anterior Segment OCT
1 Introduction
2 Proposed Method
2.1 AS-OCT Segmentation and Clinical Parameter
2.2 Multi-context Deep Network Architecture
2.3 Data Augmentation for AS-OCT
3 Experiments
4 Conclusion
References
Analysis of Morphological Changes of Lamina Cribrosa Under Acute Intraocular Pressure Change
1 Introduction
2 Methods
2.1 IOP Experiment Setup
2.2 Image Preprocessing
2.3 Unbiased Atlas Building
3 Results
3.1 Validation
3.2 Clinical Study
4 Conclusions
References
Beyond Retinal Layers: A Large Blob Detection for Subretinal Fluid Segmentation in SD-OCT Images
1 Introduction
2 Methodology
2.1 Aggregate Response Construction
2.2 Hessian Analysis
2.3 Post-pruning
3 Results
4 Conclusion
References
Automated Choroidal Neovascularization Detection for Time Series SD-OCT Images
1 Introduction
2 Method
2.1 Method Overview
2.2 Preprocessing
2.3 Classification of positive and negative samples
2.4 3D-HOG Feature Extraction
2.5 Similarity measurement and model update
2.6 Post-processing
3 Experiments
3.1 Quantitative Evaluation
3.2 Qualitative Analysis
4 Conclusions
References
CapsDeMM: Capsule Network for Detection of Munro's Microabscess in Skin Biopsy Images
1 Introduction
2 Proposed Methodology
3 Dataset
4 Experiments
4.1 Experimental Setting
4.2 Results and Discussion
5 Conclusion and Future Work
References
Webly Supervised Learning for Skin Lesion Classification
1 Introduction
2 Methodology
2.1 Model Learning
3 Experiments
4 Results and Discussion
5 Conclusions
References
Feature Driven Local Cell Graph (FeDeG): Predicting Overall Survival in Early Stage Lung Cancer
Abstract
1 Introduction
2 Brief Overview and Novel Contributions
3 Feature Driven Cell Graph
3.1 Nuclei Segmentation and Morphologic Feature Extraction
3.2 FeDeG Construction in Nuclear Morphologic Feature Space
3.3 FeDeG Features Computation
4 Experimental Design
4.1 Dataset Description
4.2 Comparative Methods
4.2.1 Graph Based and Other Pathomic Strategies
4.2.2 Deep Learning
5 Results and Discussion
5.1 Discrimination of Different Graph and Deep Learning Representations
5.2 Survival Analysis
6 Concluding Remarks
References
Cardiac, Chest and Abdominal Applications: Cardiac Imaging Applications
Towards Accurate and Complete Registration of Coronary Arteries in CTA Images
1 Introduction
2 Methods
2.1 Bifurcation Matching
2.2 Segment Registration
2.3 Further Segmentation and Registration
3 Experiments and Results
4 Discussion and Conclusion
References
Quantifying Tensor Field Similarity with Global Distributions and Optimal Transport
1 Introduction
2 Background
3 Quantifying Tensor Field Similarity
4 Experiments and Results
5 Discussion
References
Cardiac Motion Scoring with Segment- and Subject-Level Non-local Modeling
1 Introduction
2 Cardiac Motion Scoring
2.1 Myocardium Segment Re-sampling in Polar Coordinate System
2.2 Motion Scoring Neural Network
3 Dataset and Configurations
4 Results and Analysis
5 Conclusions
References
Computational Heart Modeling for Evaluating Efficacy of MRI Techniques in Predicting Appropriate ICD Therapy
Abstract
1 Introduction
2 Methods
2.1 Study Subjects and Image Acquisition
2.2 Image Processing
2.3 Simulation of Cardiac Electrophysiology
3 Results
4 Discussion and Conclusion
References
Multiview Two-Task Recursive Attention Model for Left Atrium and Atrial Scars Segmentation
1 Introduction
2 Method
3 Experimental Results and Discussion
4 Conclusion
References
Learning Interpretable Anatomical Features Through Deep Generative Models: Application to Cardiac Remodeling
1 Introduction
1.1 Related Work
2 Materials and Methods
2.1 Datasets
2.2 Deep Generative Model
3 Results
4 Discussion and Conclusion
References
Joint Learning of Motion Estimation and Segmentation for Cardiac MR Image Sequences
1 Introduction
1.1 Related Work
2 Methods
2.1 Unsupervised Cardiac Motion Estimation
2.2 Joint Model for Cardiac Motion Estimation and Segmentation
3 Experiments and Results
4 Conclusion
References
Multi-Input and Dataset-Invariant Adversarial Learning (MDAL) for Left and Right-Ventricular Coverage Estimation in Cardiac MRI
1 Introduction
2 Methodology
2.1 Problem Formulation
2.2 Multi-Input and Dataset-Invariant Adversarial Learning
2.3 Optimization
2.4 Detection and Regression for Basal/Apical Slice Position
3 Experiments and Analysis
4 Conclusion
References
Factorised Spatial Representation Learning: Application in Semi-supervised Myocardial Segmentation
1 Introduction
2 Related Work
3 Proposed Approach: The SDNet
4 Experiments and Discussion
4.1 Data and Baselines
4.2 Latent Space Arithmetic
4.3 Semi-supervised Results
5 Conclusion
References
High-Dimensional Bayesian Optimization of Personalized Cardiac Model Parameters via an Embedded Generative Model
1 Introduction
2 Background: Cardiac Electrophysiological System
3 HD Parameter Estimation
3.1 LD-to-HD Parameter Generation via VAE
3.2 Bayesian Optimization with Embedded Generative Model
4 Experiments
References
Generative Modeling and Inverse Imaging of Cardiac Transmembrane Potential
1 Introduction
2 Generative Modeling of TMP via Sequential VAE
3 Transmural EP Imaging
4 Results
5 Discussion and Conclusions:
References
Pulmonary Vessel Tree Matching for Quantifying Changes in Vascular Morphology
1 Introduction
2 Methods
2.1 Vascular Tree Construction
2.2 Vascular Tree Matching
2.3 Quantitative Analysis
3 Experiment
4 Results
5 Discussion and Conclusion
References
MuTGAN: Simultaneous Segmentation and Quantification of Myocardial Infarction Without Contrast Agents via Joint Adversarial Learning
1 Introduction
2 Methodology
2.1 MuTGAN Formulation
2.2 Generator
2.3 Discriminator
3 Materials and Implementation Details
4 Experiments and Results
5 Conclusions
References
More Knowledge Is Better: Cross-Modality Volume Completion and 3D+2D Segmentation for Intracardiac Echocardiography Contouring
1 Introduction
2 Method
2.1 3D Sparse Volume Segmentation and Completion
2.2 2D Contour Refinement
3 Experiments
4 Conclusions and Future Work
References
Unsupervised Domain Adaptation for Automatic Estimation of Cardiothoracic Ratio
1 Introduction
2 Methodology
2.1 Adversarial Training for Supervised Semantic Segmentation
2.2 Unsupervised Domain Adaption
2.3 Estimation of CTR
2.4 Semi-Supervised Semantic Segmentation
3 Experimental Results
4 Conclusions
References
TextRay: Mining Clinical Reports to Gain a Broad Understanding of Chest X-Rays
1 Introduction
2 Materials and Methods
2.1 Model
2.2 Evaluation Sets
3 Results
4 Discussion
5 Conclusion
References
Localization and Labeling of Posterior Ribs in Chest Radiographs Using a CRF-regularized FCN with Local Refinement
1 Introduction
2 Method
2.1 Generating Localization Hypotheses Using a U-Net
2.2 Selecting Reasonable Localization Hypotheses Using a CRF
2.3 Going Beyond Potentially Incorrect Localization Hypotheses
3 Experiments and Results
4 Discussion and Conclusions
References
Evaluation of Collimation Prediction Based on Depth Images and Automated Landmark Detection for Routine Clinical Chest X-Ray Exams
Abstract
1 Introduction
2 Methods
2.1 Data Pre-processing
2.2 Annotation of Reference Lung Landmarks and Lung Bounding Box
2.3 Trained Classifier for the Detection of Anatomical Landmarks
2.4 Statistical Models for the Computation of Collimation Parameters
3 Results
4 Discussion/Conclusions
References
Efficient Active Learning for Image Classification and Segmentation Using a Sample Selection and Conditional Generative Adversarial Network
1 Introduction
2 Methods
2.1 Conditional Generative Adversarial Networks
2.2 Sample Informativeness Using Uncertainty Form Bayesian Neural Networks
2.3 Implementation Details
3 Experiments
3.1 Classification Results
3.2 Segmentation Performance
3.3 Savings in Annotation Effort
4 Conclusion
References
Iterative Attention Mining for Weakly Supervised Thoracic Disease Pattern Localization in Chest X-Rays
1 Introduction
2 Methods
2.1 Disease Pattern Localization with Attention Mining
2.2 Incremental Learning with Knowledge Preservation
2.3 Multi-scale Aggregation
3 Experimental Results and Analysis
3.1 Multiple Scale Aggregation
3.2 Disease Pattern Localization with AM and KP
4 Conclusion
References
Task Driven Generative Modeling for Unsupervised Domain Adaptation: Application to X-ray Image Segmentation
1 Introduction
2 Methodology
2.1 Problem Overview
2.2 Dense Image to Image Network for Segmentation on DRRs
2.3 Task Driven Generative Adversarial Networks (TD-GAN)
3 Experiments and Results
4 Discussions and Conclusions
References
Cardiac, Chest and Abdominal Applications: Colorectal, Kidney and Liver Imaging Applications
Towards Automated Colonoscopy Diagnosis: Binary Polyp Size Estimation via Unsupervised Depth Learning
1 Introduction
2 Methods
2.1 Spatio-temporal Video Based Polyp Detection
2.2 Two-Category Polyp Size Estimation
3 Experimental Results
3.1 Polyp Detection
3.2 Polyp Size Estimation
4 Discussion
5 Conclusions
References
RIIS-DenseNet: Rotation-Invariant and Image Similarity Constrained Densely Connected Convolutional Network for Polyp Detection
1 Introduction
2 Method: RIIS-DenseNet Model
2.1 Data Rotation Augmentation
2.2 DenseNet
2.3 Joint Loss Function
3 Experiment Setup and Results
4 Conclusion
References
Interaction Techniques for Immersive CT Colonography: A Professional Assessment
1 Introduction
2 Immersive CT Colonography System
2.1 Apparatus
2.2 3D Data
2.3 Interaction Design
3 Evaluation with Professionals
4 Conclusions
References
Quasi-automatic Colon Segmentation on T2-MRI Images with Low User Effort
1 Introduction
2 Method Overview
2.1 Tubularity Detection Filter
3 Evaluation and Results
4 Conclusions
References
Ordinal Multi-modal Feature Selection for Survival Analysis of Early-Stage Renal Cancer
1 Introduction
2 Method
3 Experimental Results
4 Conclusion
References
Noninvasive Determination of Gene Mutations in Clear Cell Renal Cell Carcinoma Using Multiple Instance Decisions Aggregated CNN
1 Introduction
2 Materials and Methods
2.1 Data
2.2 Multiple Instance Decision Aggregation for Mutation Detection
3 Results
4 Conclusions
References
Combining Convolutional and Recurrent Neural Networks for Classification of Focal Liver Lesions in Multi-phase CT Images
Abstract
1 Introduction
2 Methodology
2.1 ResGLNet
2.2 BD-LSTM
2.3 Combining ResGLNet and BD-LSTM
2.4 Training Strategy
2.5 Post-processing of Label Map and Classification of Lesions
3 Experiments
3.1 Data and Implementation
3.2 Results
4 Conclusions
Acknowledgements
Acknowledgements
References
Construction of a Spatiotemporal Statistical Shape Model of Pediatric Liver from Cross-Sectional Data
1 Introduction
2 Methods
2.1 PCA with Temporal Regularization
2.2 Data Augmentation
3 Experiments
4 Conclusion
References
Deep 3D Dose Analysis for Prediction of Outcomes After Liver Stereotactic Body Radiation Therapy
1 Introduction
2 Methodology
2.1 Registration of 3D Dose Plans into Anatomy-Unified Space
2.2 CNN Transfer Learning for Outcome Prediction After Liver SBRTs
2.3 Treatment Outcome Atlases
3 Experiments and Results
3.1 Post-SBRT Survival and Local Progression Prediction
3.2 Survival and Local Progression Atlases
4 Discussion and Conclusion
References
Liver Lesion Detection from Weakly-Labeled Multi-phase CT Volumes with a Grouped Single Shot MultiBox Detector
1 Introduction
2 Multi-phase Data
3 Grouped Single Shot MultiBox Detector
4 Experiments
5 Discussion and Conclusions
References
A Diagnostic Report Generator from CT Volumes on Liver Tumor with Semi-supervised Attention Mechanism
1 Introduction
2 Model
3 Experiments
4 Conclusion
References
Less is More: Simultaneous View Classification and Landmark Detection for Abdominal Ultrasound Images
1 Introduction
2 Methods and Results
2.1 Data and Task Definitions
2.2 MTL Framework
2.3 Results
3 Discussion
References
Cardiac, Chest and Abdominal Applications: Lung Imaging Applications
Deep Active Self-paced Learning for Accurate Pulmonary Nodule Segmentation
1 Introduction
2 Method
2.1 Nodule R-CNN
2.2 The Deep Active Self-paced Learning Strategy
3 Experiments and Results
4 Conclusions
References
CT-Realistic Lung Nodule Simulation from 3D Conditional Generative Adversarial Networks for Robust Lung Segmentation
1 Introduction
2 Methods
2.1 CGAN Formulation
2.2 3D CGAN Architecture
2.3 CGAN Optimization
3 Experiments and Results
3.1 3D CGAN Performance
3.2 Improving Pathological Lung Segmentation
4 Conclusion
References
Fast CapsNet for Lung Cancer Screening
1 Introduction
2 Capsule Network
3 Fast Capsule Network
4 Experiments
5 Conclusions
References
Mean Field Network Based Graph Refinement with Application to Airway Tree Extraction
1 Introduction
2 Method
2.1 The Graph Refinement Model
2.2 Mean Field Network
2.3 Airway Tree Extraction as Graph Refinement
3 Experiments and Results
4 Discussion and Conclusion
References
Automated Pulmonary Nodule Detection: High Sensitivity with Few Candidates
1 Introduction
2 Methods
2.1 High Sensitivity Candidate Detection with FPN
2.2 Conditional 3D-NMS for Redundant Candidate Removal
2.3 Attention 3D-CNN for False Positive Reduction
3 Experiments
3.1 Implementation Details
3.2 Ablation Study and Results
4 Conclusion
References
Deep Learning from Label Proportions for Emphysema Quantification
1 Introduction
2 Methods
2.1 Architectures
2.2 A Loss for Learning from Proportion Intervals (LPI)
3 Experimental Setting
4 Results
5 Discussion and Conclusion
References
Tumor-Aware, Adversarial Domain Adaptation from CT to MRI for Lung Cancer Segmentation
1 Introduction
2 Method
2.1 Step 1: MRI Synthesis Using Tumor-Aware Unsupervised Cross Domain Adaptation
2.2 Step 2: Semi-supervised Tumor Segmentation from MRI
2.3 Network Structure and Implementation
3 Experiments and Results
3.1 Ablation Tests
3.2 Datasets
3.3 MR Image Synthesis Results
3.4 Segmentation Results
4 Discussion
5 Conclusions
References
From Local to Global: A Holistic Lung Graph Model
1 Introduction
2 Methods
2.1 Dataset
2.2 Holistic Graph Model of the Lungs
2.3 Undirected Weighted Graph Model of the Lung
2.4 Directed Weighted Graph-Model of the Lungs
2.5 Graph-Based Patient Descriptor
3 Experimental Setup
4 Results
5 Discussion
6 Conclusions and Future Work
References
S4ND: Single-Shot Single-Scale Lung Nodule Detection
1 Introduction
2 Method
2.1 Single-Scale Detection
2.2 Dense and Deeper Convolution Blocks Improve Detection
2.3 Max-Pooling Improves Detection
2.4 Proposed 3D Deep Network Architecture
3 Experiments and Results
4 Conclusion
References
Vascular Network Organization via Hough Transform (VaNgOGH): A Novel Radiomic Biomarker for Diagnosis and Treatment Response
1 Introduction
2 Previous Work and Novel Contributions
3 Methodology
3.1 Notation
3.2 VaNgOGH Descriptor
4 Experimental Results and Discussion
4.1 Data Description
4.2 Comparative Strategies and Classifier Construction
4.3 Experiment 1: Pre-treatment Response Prediction in Breast Cancer DCE-MRI
4.4 Experiment 2: Malignancy Diagnosis for Lung Nodules
5 Concluding Remarks
References
DeepEM: Deep 3D ConvNets with EM for Weakly Supervised Pulmonary Nodule Detection
1 Introduction
2 DeepEM for Weakly Supervised Detection
3 Experiments
4 Conclusion
References
Statistical Framework for the Definition of Emphysema in CT Scans: Beyond Density Mask
1 Introduction
2 Characterization of Emphysema in CT Scans
3 Adaptive Threshold for Emphysema Detection
4 Results
5 Conclusion
References
Cardiac, Chest and Abdominal Applications: Breast Imaging Applications
Conditional Generative Adversarial and Convolutional Networks for X-ray Breast Mass Segmentation and Shape Classification
1 Introduction
2 Related Work
3 Proposed Model
3.1 System Overview
3.2 Mass Segmentation Model (with cGAN)
3.3 Shape classification model (with CNN)
4 Experiments
4.1 Datasets
4.2 Experimental Results
5 Conclusions
References
A Robust and Effective Approach Towards Accurate Metastasis Detection and pN-stage Classification in Breast Cancer
1 Introduction
2 Methodology
2.1 Regions of Interests Extraction
2.2 Metastasis Detection
2.3 Lymph Node Classification
3 Experiments
3.1 Dataset
3.2 Evaluation Metrics
3.3 Experimental Details
3.4 Results
4 Conclusion
References
3D Anisotropic Hybrid Network: Transferring Convolutional Features from 2D Images to 3D Anisotropic Volumes
1 Introduction
2 Anisotropic Hybrid Network
2.1 Pre-training a Multi-Channel 2D Feature Encoder
2.2 Transferring the Learned 2D Features into 3D AH-Net
2.3 Anisotropic Hybrid Decoder
2.4 Training the AH-Net
3 Experimental Results
3.1 Breast Lesion Detection from DBT
3.2 Liver and Liver Tumor Segmentation from CT
4 Conclusion
References
Deep Generative Breast Cancer Screening and Diagnosis
1 Introduction
2 Methods
2.1 Model Overview
2.2 GAN-Enhanced Deep Classification
2.3 Training
2.4 Transfer Learning
3 Experiments and Results
3.1 Performance Enhanced by GAN
4 Conclusion
References
Integrate Domain Knowledge in Training CNN for Ultrasonography Breast Cancer Diagnosis
Abstract
1 Introduction
2 Method
2.1 Data Acquisition
2.2 Image Preprocessing
2.3 Training Multi-task CNN
3 Results
3.1 Classification Result
3.2 Examples of Correct and Wrong Predictions
4 Conclusion
Acknowledgement
References
Small Lesion Classification in Dynamic Contrast Enhancement MRI for Breast Cancer Early Detection
1 Introduction
2 Methods
2.1 Datasets
2.2 DC-LSTM
2.3 Latent Attributes Learning
3 Experiments
3.1 DC-LSTM
3.2 Latent Attributes Learning
4 Conclusion
References
Thermographic Computational Analyses of a 3D Model of a Scanned Breast
1 Introduction
2 Materials and Methods
2.1 Heat Biotransference Equation
2.2 3D Scanned Computational Model and Numerical Solution
2.3 Case of Simulated Breast Cancer
3 Results and Discussions
4 Conclusions
References
Y-Net: Joint Segmentation and Classification for Diagnosis of Breast Biopsy Images
1 Introduction
2 A System for Joint Segmentation and Classification
2.1 Y-Net Architecture
2.2 Discriminative Instance Selection
2.3 Diagnostic Classification
3 Experiments
3.1 Segmentation Results
3.2 Diagnostic Classification Results
4 Conclusion
References
Cardiac, Chest and Abdominal Applications: Other Abdominal Applications
AutoDVT: Joint Real-Time Classification for Vein Compressibility Analysis in Deep Vein Thrombosis Ultrasound Diagnostics
1 Introduction
2 Method
3 Data and Experiments
4 Results and Discussion
References
MRI Measurement of Placental Perfusion and Fetal Blood Oxygen Saturation in Normal Pregnancy and Placental Insufficiency
1 Introduction
2 Methods
3 Results
4 Discussion
References
Automatic Lacunae Localization in Placental Ultrasound Images via Layer Aggregation
1 Introduction
2 Methods
2.1 From Dot Annotation to Confidence Map
2.2 Lacunae Localization: Layer Aggregation Approaches
3 Experiments
4 Discussion
References
A Decomposable Model for the Detection of Prostate Cancer in Multi-parametric MRI
1 Introduction
2 Related Works
3 Methods
3.1 Random Ferns
3.2 Random Fern Decomposition
3.3 Data
3.4 Prostate Segmentation
3.5 Training
3.6 Statistical Analysis
4 Results
5 Discussion
6 Conclusion
References
Direct Automated Quantitative Measurement of Spine via Cascade Amplifier Regression Network
1 Introduction
2 Cascade Amplifier Regression Network
2.1 Mathematical Formulation
2.2 CARN Architecture
2.3 Local Shape-Constrained Manifold Regularization Loss Function
3 Experimental Results
4 Conclusions
References
Estimating Achilles Tendon Healing Progress with Convolutional Neural Networks
1 Introduction
2 Method
3 Experiments
3.1 Dataset
3.2 Tendon Healing Process Assessment
4 Conclusions
References
Author Index