E-Book, Englisch, 452 Seiten
Lu / Wang / Carneiro Deep Learning and Convolutional Neural Networks for Medical Imaging and Clinical Informatics
1. Auflage 2019
ISBN: 978-3-030-13969-8
Verlag: Springer International Publishing
Format: PDF
Kopierschutz: 1 - PDF Watermark
E-Book, Englisch, 452 Seiten
Reihe: Advances in Computer Vision and Pattern Recognition
ISBN: 978-3-030-13969-8
Verlag: Springer International Publishing
Format: PDF
Kopierschutz: 1 - PDF Watermark
This book reviews the state of the art in deep learning approaches to high-performance robust disease detection, robust and accurate organ segmentation in medical image computing (radiological and pathological imaging modalities), and the construction and mining of large-scale radiology databases. It particularly focuses on the application of convolutional neural networks, and on recurrent neural networks like LSTM, using numerous practical examples to complement the theory.
The book's chief features are as follows: It highlights how deep neural networks can be used to address new questions and protocols, and to tackle current challenges in medical image computing; presents a comprehensive review of the latest research and literature; and describes a range of different methods that employ deep learning for object or landmark detection tasks in 2D and 3D medical imaging. In addition, the book examines a broad selection of techniques for semantic segmentation using deep learning principles in medical imaging; introduces a novel approach to text and image deep embedding for a large-scale chest x-ray image database; and discusses how deep learning relational graphs can be used to organize a sizable collection of radiology findings from real clinical practice, allowing semantic similarity-based retrieval.The intended reader of this edited book is a professional engineer, scientist or a graduate student who is able to comprehend general concepts of image processing, computer vision and medical image analysis. They can apply computer science and mathematical principles into problem solving practices. It may be necessary to have a certain level of familiarity with a number of more advanced subjects: image formation and enhancement, image understanding, visual recognition in medical applications, statistical learning, deep neural networks, structured prediction and image segmentation.
Dr. Le Lu is the Director of Ping An Technology US Research Labs, and an adjunct faculty member at Johns Hopkins University, USA. Dr. Xiaosong Wang is a Senior Applied Research Scientist at Nvidia Corp., USA. Dr. Gustavo Carneiro is an Associate Professor at the University of Adelaide, Australia. Dr. Lin Yang is an Associate Professor at the University of Florida, USA.
Autoren/Hrsg.
Weitere Infos & Material
1;Preface;6
1.1;Organization and Features;7
2;Contents;9
3;Part I Segmentation;12
4;1 Pancreas Segmentation in CT and MRI via Task-Specific Network Design and Recurrent Neural Contextual Learning;13
4.1;1.1 Introduction;14
4.2;1.2 Convolutional Neural Network for Pancreas Segmentation;15
4.2.1;1.2.1 Design of Network Architecture;15
4.2.2;1.2.2 Design of Model Training Strategy;16
4.2.3;1.2.3 Design of Loss Functions;18
4.2.4;1.2.4 Experimental Results;19
4.3;1.3 Recurrent Neural Network for Contextual Learning;22
4.4;1.4 Recurrent Neural Network;23
4.4.1;1.4.1 Bidirectional Contextual Regularization;25
4.4.2;1.4.2 Experimental Results;25
4.5;1.5 State-of-the-Art Methods for Pancreas Segmentation;27
4.6;1.6 Summary;29
4.7;References;30
5;2 Deep Learning for Muscle Pathology Image Analysis;32
5.1;2.1 Introduction;33
5.2;2.2 Muscle Perimysium Segmentation;33
5.2.1;2.2.1 Recurrent Neural Network;34
5.3;2.3 WSI Inflammatory Muscle Disease Subtype Classification;40
5.3.1;2.3.1 Methodology;40
5.4;2.4 Experimental Results;43
5.4.1;2.4.1 Dataset;43
5.4.2;2.4.2 Implementation Details;43
5.4.3;2.4.3 Evaluation of Different WSI Frameworks;44
5.4.4;2.4.4 Evaluation of Different Training Methods;46
5.4.5;2.4.5 Evaluation of Different Number of ROIs;46
5.4.6;2.4.6 Diagnosis Interpretation and Visualization;47
5.5;2.5 Summary;48
5.6;References;49
6;3 2D-Based Coarse-to-Fine Approaches for Small Target Segmentation in Abdominal CT Scans;51
6.1;3.1 Introduction;52
6.2;3.2 Related Work;53
6.3;3.3 A Step-Wise Coarse-to-Fine Approach for Medical Image Segmentation;54
6.3.1;3.3.1 Deep Segmentation Networks;55
6.3.2;3.3.2 Fixed-Point Optimization;55
6.3.3;3.3.3 Application to Pancreatic Cyst Segmentation;58
6.4;3.4 An End-to-End Coarse-to-Fine Approach for Medical Image Segmentation;60
6.4.1;3.4.1 Recurrent Saliency Transformation Network;60
6.4.2;3.4.2 Training and Testing;62
6.4.3;3.4.3 Application to Pancreatic Cyst Segmentation;63
6.5;3.5 Pancreas Segmentation Experiments;64
6.5.1;3.5.1 Dataset and Evaluation;64
6.5.2;3.5.2 Evaluation of the Step-Wise Coarse-to-Fine Approach;64
6.5.3;3.5.3 Evaluation of the End-to-End Coarse-to-Fine Approach;66
6.6;3.6 JHMI Multi-organ Segmentation Experiments;70
6.7;3.7 JHMI Pancreatic Cyst Segmentation Experiments;71
6.8;3.8 Conclusions;72
6.9;References;72
7;4 Volumetric Medical Image Segmentation: A 3D Deep Coarse-to-Fine Framework and Its Adversarial Examples;76
7.1;4.1 Introduction;77
7.2;4.2 Related Work;79
7.2.1;4.2.1 Deep Learning-Based Medical Image Segmentation;79
7.2.2;4.2.2 Adversarial Attacks and Defenses for Medical Image Segmentation Networks;81
7.3;4.3 Method;81
7.3.1;4.3.1 A 3D Coarse-to-Fine Framework for Medical Image Segmentation;81
7.3.2;4.3.2 3D Adversarial Examples;86
7.4;4.4 Experiments;87
7.4.1;4.4.1 Pancreas Segmentation;88
7.4.2;4.4.2 Adversarial Attack and Defense;93
7.5;4.5 Conclusion;96
7.6;References;96
8;5 Unsupervised Domain Adaptation of ConvNets for Medical Image Segmentation via Adversarial Learning;99
8.1;5.1 Introduction;100
8.2;5.2 Related Works;101
8.2.1;5.2.1 Feature-Level Adaptation;101
8.2.2;5.2.2 Pixel-Level Adaptation;102
8.3;5.3 Feature-Level Adaptation with Latent Space Alignment;103
8.3.1;5.3.1 Method;103
8.3.2;5.3.2 Experimental Results;107
8.4;5.4 Pixel-Level Adaptation with Image-to-Image Translation;112
8.4.1;5.4.1 Method;112
8.4.2;5.4.2 Experimental Results;115
8.5;5.5 Discussion;118
8.6;5.6 Conclusion;119
8.7;References;119
9;Part II Detection and Localization;122
10;6 Glaucoma Detection Based on Deep Learning Network in Fundus Image;123
10.1;6.1 Introduction;124
10.2;6.2 M-Net: Multi-label Segmentation Network;126
10.2.1;6.2.1 Multi-scale U-Shape Network;126
10.2.2;6.2.2 Side-Output Layer;128
10.2.3;6.2.3 Multi-label Loss Function;128
10.2.4;6.2.4 Polar Transformation;129
10.3;6.3 DENet: Disc-Aware Ensemble Network;130
10.3.1;6.3.1 Global Fundus Image Level;131
10.3.2;6.3.2 Optic Disc Region Level;132
10.4;6.4 Experiments;133
10.4.1;6.4.1 Implementation;133
10.4.2;6.4.2 Segmentation Evaluation;133
10.4.3;6.4.3 Glaucoma Screening Evaluation;136
10.4.4;6.4.4 REFUGE Challenge;138
10.5;6.5 Conclusion;139
10.6;References;140
11;7 Thoracic Disease Identification and Localization with Limited Supervision;142
11.1;7.1 Introduction;143
11.2;7.2 Related Work;145
11.3;7.3 Model;146
11.3.1;7.3.1 Image Model;146
11.3.2;7.3.2 Loss Function;147
11.3.3;7.3.3 Localization Generation;149
11.3.4;7.3.4 Training;150
11.4;7.4 Experiments;151
11.4.1;7.4.1 Disease Identification;151
11.4.2;7.4.2 Disease Localization;154
11.4.3;7.4.3 Qualitative Results;161
11.5;7.5 Conclusion;162
11.6;References;162
12;8 Deep Reinforcement Learning for Detecting Breast Lesions from DCE-MRI;165
12.1;8.1 Introduction;166
12.2;8.2 Literature Review;168
12.3;8.3 Methods;169
12.3.1;8.3.1 Data Set;169
12.3.2;8.3.2 Detection Method;170
12.3.3;8.3.3 Training;171
12.3.4;8.3.4 Inference;173
12.4;8.4 Experiments;173
12.4.1;8.4.1 Data Set;174
12.4.2;8.4.2 Experimental Setup;174
12.4.3;8.4.3 Experimental Results;175
12.5;8.5 Discussion;177
12.6;8.6 Conclusion and Future Work;177
12.7;References;178
13;9 Automatic Vertebra Labeling in Large-Scale Medical Images Using Deep Image-to-Image Network with Message Passing and Sparsity Regularization;181
13.1;9.1 Introduction;182
13.2;9.2 Methodology;185
13.2.1;9.2.1 The Deep Image-to-Image Network (DI2IN) for Spinal Centroid Localization;185
13.2.2;9.2.2 Probability Map Enhancement with Message Passing;187
13.2.3;9.2.3 Joint Refinement Using Shape-Based Dictionaries;189
13.3;9.3 Experiments;192
13.4;9.4 Conclusion;197
13.5;References;197
14;10 Anisotropic Hybrid Network for Cross-Dimension Transferable Feature Learning in 3D Medical Images;200
14.1;10.1 Introduction;201
14.2;10.2 Related Work;202
14.3;10.3 Anisotropic Hybrid Network;203
14.3.1;10.3.1 Learning a Multichannel 2D Feature Encoder;204
14.3.2;10.3.2 Transferring the Learned 2D Net to 3D AH-Net;205
14.3.3;10.3.3 Anisotropic Hybrid Decoder;208
14.4;10.4 Experimental Results;208
14.4.1;10.4.1 Breast Lesion Detection from DBT;209
14.4.2;10.4.2 Liver and Liver Tumor Segmentation from CT;213
14.5;10.5 Conclusion;216
14.6;References;216
15;Part III Various Applications;218
16;11 Deep Hashing and Its Application for Histopathology Image Analysis;219
16.1;11.1 Introduction;219
16.2;11.2 Deep Hashing;220
16.2.1;11.2.1 Pointwise-Based Hashing;220
16.2.2;11.2.2 Multiwise-Based Hashing;221
16.2.3;11.2.3 Pairwise-Based Hashing;222
16.3;11.3 Experimental Results and Discussion;229
16.3.1;11.3.1 Experimental Results;231
16.3.2;11.3.2 Discussion;233
16.4;11.4 Summary;234
16.5;References;235
17;12 Tumor Growth Prediction Using Convolutional Networks;238
17.1;12.1 Introduction;239
17.2;12.2 Group Learning Approach for Tumor Growth Prediction;240
17.2.1;12.2.1 Image Processing and Patch Extraction;242
17.2.2;12.2.2 Learning a Voxel-Wise Deep Representation;243
17.2.3;12.2.3 Learning a Predictive Model with Multi-source Features;244
17.2.4;12.2.4 Experiments and Results;245
17.3;12.3 Convolutional Invasion and Expansion Networks for Tumor Growth Prediction;247
17.3.1;12.3.1 Learning Invasion Network;248
17.3.2;12.3.2 Learning Expansion Network;249
17.3.3;12.3.3 Fusing Invasion and Expansion Networks;250
17.3.4;12.3.4 Personalizing Invasion and Expansion Networks;252
17.3.5;12.3.5 Predicting with Invasion and Expansion Networks;253
17.3.6;12.3.6 Experimental Methods and Results;253
17.4;12.4 Summary;257
17.5;References;257
18;13 Deep Spatial-Temporal Convolutional Neural Networks for Medical Image Restoration;260
18.1;13.1 Introduction;261
18.2;13.2 Related Work;262
18.2.1;13.2.1 Radiation Dose Reduction;263
18.2.2;13.2.2 Image Restoration;263
18.2.3;13.2.3 Spatial-Temporal Architecture;264
18.3;13.3 Methodology;265
18.3.1;13.3.1 Spatial-Temporal Patches;265
18.3.2;13.3.2 Deep Spatial-Temporal Network;265
18.4;13.4 Platform and Data Acquisition;267
18.4.1;13.4.1 Computational Platform;267
18.4.2;13.4.2 Datasets;267
18.4.3;13.4.3 Low Radiation Dose Simulation and Data Preprocessing;268
18.5;13.5 Experiments and Results;269
18.5.1;13.5.1 Evaluation Metrics;269
18.5.2;13.5.2 Spatial-Temporal Super-Resolution and Denoising;270
18.6;13.6 Conclusion;272
18.7;References;273
19;14 Generative Low-Dose CT Image Denoising;275
19.1;14.1 Introduction;276
19.2;14.2 Methods;279
19.2.1;14.2.1 Noise Reduction Model;279
19.2.2;14.2.2 WGAN;279
19.2.3;14.2.3 Perceptual Loss;280
19.2.4;14.2.4 Network Structures;280
19.2.5;14.2.5 Other Networks;282
19.3;14.3 Experiments;282
19.3.1;14.3.1 Experimental Datasets;282
19.3.2;14.3.2 Network Training;283
19.3.3;14.3.3 Network Convergence;283
19.3.4;14.3.4 Denoising Results;285
19.3.5;14.3.5 Quantitative Analysis;287
19.4;14.4 Discussions and Conclusion;291
19.5;References;293
20;15 Image Quality Assessment for Population Cardiac Magnetic Resonance Imaging;296
20.1;15.1 Introduction;297
20.2;15.2 Full LV Coverage Detection Method;300
20.2.1;15.2.1 Problem Formulation;300
20.2.2;15.2.2 Dataset Invariance 3D Intensity Representations;301
20.2.3;15.2.3 Fisher Discriminative 3D CNN Model;304
20.3;15.3 Materials and Metrics;306
20.3.1;15.3.1 CMR Acquisition Protocol and Annotation;306
20.3.2;15.3.2 Training and Testing Set Definitions;307
20.3.3;15.3.3 Learning Performance Metrics;309
20.4;15.4 Experiments and Results;309
20.4.1;15.4.1 Hyper Parameter Selection on UK Biobank;309
20.4.2;15.4.2 Dataset Adversarial Learning Performance;313
20.4.3;15.4.3 Intra-rater Agreement of Full LV Coverage Detection;315
20.4.4;15.4.4 Implementation Considerations;315
20.5;15.5 Conclusion;316
20.6;References;316
21;16 Agent-Based Methods for Medical Image Registration;319
21.1;16.1 Introduction;319
21.2;16.2 Background;320
21.2.1;16.2.1 Parametric Image Registration;321
21.2.2;16.2.2 Image Registration Using Deep Learning;321
21.2.3;16.2.3 Deep Reinforcement Learning;322
21.2.4;16.2.4 Special Euclidean Group SE(3);322
21.3;16.3 Agent-Based Image Registration;322
21.3.1;16.3.1 Image Registration as an MDP;323
21.3.2;16.3.2 Action Space;324
21.3.3;16.3.3 Reward System;325
21.3.4;16.3.4 Agent Observation;325
21.3.5;16.3.5 Learning Policy with Supervised Learning;327
21.3.6;16.3.6 Multi-agent System;328
21.4;16.4 Agent-Based 3-D/3-D Image Registration;329
21.4.1;16.4.1 Implementation;329
21.4.2;16.4.2 Experiments and Results;331
21.5;16.5 Agent-Based 2-D/3-D Image Registration;333
21.5.1;16.5.1 Implementation;334
21.5.2;16.5.2 Experiments and Results;336
21.6;16.6 Discussion;339
21.7;References;340
22;17 Deep Learning for Functional Brain Connectivity: Are We There Yet?;342
22.1;17.1 Introduction;343
22.2;17.2 Related Work;344
22.3;17.3 Methods;346
22.3.1;17.3.1 fMRI Preprocessing and Feature Extraction;346
22.3.2;17.3.2 Ensemble Classification Approach;347
22.3.3;17.3.3 Deep Learning Models;348
22.4;17.4 Results;351
22.4.1;17.4.1 Independent Classifiers;351
22.4.2;17.4.2 Ensemble Classifiers;351
22.4.3;17.4.3 Deep Learning Classifiers;353
22.5;17.5 Discussion and Conclusions;355
22.6;References;358
23;Part IV Large-Scale Data Mining and Data Synthesis;361
24;18 ChestX-ray: Hospital-Scale Chest X-ray Database and Benchmarks on Weakly Supervised Classification and Localization of Common Thorax Diseases;362
24.1;18.1 Introduction;363
24.1.1;18.1.1 Recent Advances;365
24.2;18.2 Database Construction;366
24.2.1;18.2.1 Disease Label Mining;366
24.2.2;18.2.2 Evaluation on Mined Disease Labels;369
24.2.3;18.2.3 Chest X-ray Image Processing and Hand-Labeled Ground Truth;370
24.3;18.3 Applications on Constructed Database;371
24.3.1;18.3.1 Classification and Localization Framework;371
24.4;18.4 Evaluations;374
24.5;18.5 Extension to 14 Common Thorax Disease Labels;377
24.5.1;18.5.1 Evaluation of NLP Mined Labels;378
24.5.2;18.5.2 Benchmark Results;378
24.6;18.6 Summary;381
24.7;References;382
25;19 Automatic Classification and Reporting of Multiple Common Thorax Diseases Using Chest Radiographs;386
25.1;19.1 Introduction;387
25.2;19.2 Previous Works in CAD;389
25.3;19.3 Multi-level Attention in a Unified Framework;390
25.3.1;19.3.1 AETE: Attention on Text;390
25.3.2;19.3.2 SW-GAP: Attention on Image;392
25.3.3;19.3.3 Overall CNN-RNN Model;393
25.3.4;19.3.4 Joint Learning;394
25.4;19.4 Applications;395
25.4.1;19.4.1 Annotation of Chest X-Ray Images;395
25.4.2;19.4.2 Automatic Reporting of Thorax Diseases;395
25.5;19.5 Experiments;395
25.5.1;19.5.1 Datasets for Evaluation;395
25.5.2;19.5.2 Report Vocabulary;396
25.5.3;19.5.3 Evaluation Metrics;396
25.5.4;19.5.4 Details on Training;397
25.5.5;19.5.5 Evaluation on Image Annotation;397
25.5.6;19.5.6 Evaluation on Classification and Automated Reporting;400
25.6;19.6 Summary;403
25.7;References;403
26;20 Deep Lesion Graph in the Wild: Relationship Learning and Organization of Significant Radiology Image Findings in a Diverse Large-Scale Lesion Database;406
26.1;20.1 Introduction;407
26.2;20.2 Related Work;409
26.3;20.3 Dataset;410
26.4;20.4 Method;412
26.4.1;20.4.1 Supervision Cues;412
26.4.2;20.4.2 Learning Lesion Embeddings;414
26.4.3;20.4.3 Lesion Retrieval and Matching;416
26.5;20.5 Experiments;417
26.5.1;20.5.1 Implementation Details;418
26.5.2;20.5.2 Content-Based Lesion Retrieval;418
26.5.3;20.5.3 Intra-patient Lesion Matching;422
26.6;20.6 Conclusion and Future Work;423
26.7;References;425
27;21 Simultaneous Super-Resolution and Cross-Modality Synthesis in Magnetic Resonance Imaging;429
27.1;21.1 Introduction;430
27.2;21.2 Background;433
27.2.1;21.2.1 Image Degradation Model;433
27.2.2;21.2.2 Dictionary Learning;434
27.3;21.3 Method;434
27.3.1;21.3.1 Data Description;435
27.3.2;21.3.2 Gradient Feature Representation;436
27.3.3;21.3.3 Cross-Modality Dictionary Learning;436
27.3.4;21.3.4 Clustering-Based Globally Redundant Codes;437
27.4;21.4 Experiments;440
27.4.1;21.4.1 Dictionary Size;441
27.4.2;21.4.2 Sparsity;442
27.4.3;21.4.3 MRI Super-Resolution;443
27.4.4;21.4.4 Simultaneous Super-Resolution and Cross-Modality Synthesis;444
27.5;21.5 Conclusion;447
27.6;References;447
28;Appendix A Index;450
29;Index;450
