|
Preface |
6 |
|
|
Contents |
8 |
|
|
Part I Overview and Issues |
10 |
|
|
Deep Learning in Medical Image Analysis |
11 |
|
|
Introduction |
11 |
|
|
Deep Learning for Medical Image Analysis and CAD |
12 |
|
|
Challenges in Deep-Learning-Based CAD |
15 |
|
|
Data Collection |
17 |
|
|
Transfer Learning |
19 |
|
|
Data Augmentation |
23 |
|
|
Training, Validation, and Independent Testing |
24 |
|
|
Acceptance Testing, Preclinical Testing, and User Training |
24 |
|
|
Quality Assurance and Performance Monitoring |
25 |
|
|
Interpretability of CAD/AI Recommendations |
26 |
|
|
Summary |
26 |
|
|
Medical Image Synthesis via Deep Learning |
30 |
|
|
Introduction |
30 |
|
|
Deep Learning Models for Medical Image Synthesis |
33 |
|
|
Convolutional Neural Networks |
33 |
|
|
Generative Adversarial Networks |
34 |
|
|
Within-Modality Synthesis |
35 |
|
|
3D cGAN |
36 |
|
|
Framework |
36 |
|
|
Experimental Results |
36 |
|
|
Locality Adaptive Multi-Modality GANs |
38 |
|
|
Framework |
39 |
|
|
Experimental Results |
40 |
|
|
Cross-Modality Synthesis |
41 |
|
|
3D cGAN with Subject-Specific Local Adaptive Fusion |
41 |
|
|
Framework |
42 |
|
|
Experimental Results |
43 |
|
|
Edge-Aware GANs |
43 |
|
|
Framework |
44 |
|
|
Experimental Results |
46 |
|
|
Conclusion |
47 |
|
|
Part II Applications: Screening and Diagnosis |
52 |
|
|
Deep Learning for Pulmonary Image Analysis: Classification, Detection, and Segmentation |
53 |
|
|
Background of Lung Diseases |
53 |
|
|
Introduction |
54 |
|
|
Methods |
54 |
|
|
Classification of Lung Abnormalities |
54 |
|
|
Detection of Lung Abnormalities |
58 |
|
|
Segmentation of Lung Abnormalities |
59 |
|
|
Conclusion |
62 |
|
|
Deep Learning Computer-Aided Diagnosis for Breast Lesion in Digital Mammogram |
65 |
|
|
Introduction |
66 |
|
|
Related Work |
67 |
|
|
Materials and Methods |
68 |
|
|
Dataset |
68 |
|
|
Datasets Preparation: Training, Validation, and Testing |
69 |
|
|
Preprocessing |
69 |
|
|
Data Balancing and Augmentation |
69 |
|
|
Initialization of Trainable Parameters for Deep Learning Models |
70 |
|
|
Breast Lesion Detection via YOLO |
70 |
|
|
Breast Lesion Segmentation via FrCN |
70 |
|
|
Breast Lesion Classification via Three Convolutional Neural Networks |
71 |
|
|
Experimental Settings |
72 |
|
|
Detection Experimental Settings |
72 |
|
|
Segmentation Experimental Settings |
72 |
|
|
Classification Experimental Settings |
72 |
|
|
Implementation Environment |
73 |
|
|
Experimental Results and Discussion |
73 |
|
|
Evaluation Metrics |
73 |
|
|
Breast Lesion Detection Results |
73 |
|
|
Breast Lesion Segmentation Results |
73 |
|
|
Breast Lesion Classification Results |
75 |
|
|
Conclusion |
76 |
|
|
Decision Support System for Lung Cancer Using PET/CT and Microscopic Images |
79 |
|
|
Introduction |
79 |
|
|
Outline of Decision Support System |
80 |
|
|
Automated Detection of Lung Nodules in PET/CT Images Using Convolutional Neural Network and Radiomic Features |
81 |
|
|
Background |
81 |
|
|
Method Overview |
81 |
|
|
Initial Nodule Detection |
82 |
|
|
False Positive Reduction |
82 |
|
|
Classification Using a Convolutional Neural Network |
83 |
|
|
Handcrafted Radiomic Features |
83 |
|
|
Classification |
83 |
|
|
Results |
84 |
|
|
Image Datasets |
84 |
|
|
Evaluation Metrics |
84 |
|
|
Detection Results |
84 |
|
|
Discussion |
85 |
|
|
Automated Malignancy Analysis of Lung Nodules in PET/CT Images Using Radiomic Features |
86 |
|
|
Introduction |
86 |
|
|
Materials and Methods |
86 |
|
|
Image Dataset |
86 |
|
|
Methods Overview |
87 |
|
|
Volume of Interest (VOI) Extraction |
87 |
|
|
Extraction of Characteristic Features |
87 |
|
|
Classification |
90 |
|
|
Results |
90 |
|
|
Discussion |
90 |
|
|
Automated Malignancy Analysis Using Lung Cytological Images |
92 |
|
|
Introduction |
92 |
|
|
Materials and Methods |
93 |
|
|
Image Dataset |
93 |
|
|
Network Architecture |
93 |
|
|
Results and Discussion |
94 |
|
|
Automated Classification of Lung Cancer Types from Cytological Images |
94 |
|
|
Introduction |
94 |
|
|
Materials and Methods |
95 |
|
|
Image Dataset |
95 |
|
|
Network Architecture |
96 |
|
|
Results and Discussion |
96 |
|
|
Conclusion |
98 |
|
|
Lesion Image Synthesis Using DCGANs for Metastatic Liver Cancer Detection |
101 |
|
|
Introduction |
101 |
|
|
Proposed Method |
103 |
|
|
Dataset |
103 |
|
|
Lesion Image Generation |
103 |
|
|
Method 1: Synthesis Using Poisson Blending |
103 |
|
|
Method 2: Generation Based on a CT Value Distribution |
104 |
|
|
Method 3: Generation Using DCGANs |
105 |
|
|
Selection of the Region of Interest for Lesion Synthesis |
105 |
|
|
Detection Method |
107 |
|
|
Experiments |
108 |
|
|
Results |
109 |
|
|
Discussion |
109 |
|
|
Conclusion |
110 |
|
|
Retinopathy Analysis Based on Deep ConvolutionNeural Network |
113 |
|
|
Introduction |
113 |
|
|
General Arteriolar Narrowing Detection |
114 |
|
|
Blood Vessel Extraction |
114 |
|
|
Related Works |
114 |
|
|
Database |
115 |
|
|
Preprocessing |
115 |
|
|
Blood Vessel Extraction Using DCNN |
116 |
|
|
Detection of Arteriolar Narrowing Using AVR |
117 |
|
|
Related Works |
117 |
|
|
Database |
118 |
|
|
Classification of Arteries and Veins |
118 |
|
|
AVR Measurement |
119 |
|
|
Microaneurysm Detection |
120 |
|
|
Related Work |
120 |
|
|
Database |
121 |
|
|
Methods |
121 |
|
|
Preprocessing |
121 |
|
|
Microaneurysm Detection Based on DCNN |
122 |
|
|
Reducing the Number of False Positives |
123 |
|
|
Examination |
124 |
|
|
Conclusion |
124 |
|
|
Diagnosis of Glaucoma on Retinal Fundus Images Using Deep Learning: Detection of Nerve Fiber Layer Defect and Optic Disc Analysis |
127 |
|
|
Introduction |
127 |
|
|
Related Works |
129 |
|
|
NFLD Detection |
129 |
|
|
Background |
129 |
|
|
Proposed Method |
130 |
|
|
Segmentation Network |
130 |
|
|
Detection Network |
131 |
|
|
Combined Method |
131 |
|
|
Dataset |
131 |
|
|
Preprocessing |
133 |
|
|
Evaluation |
133 |
|
|
Results |
133 |
|
|
Discussion |
133 |
|
|
Optic Disc Analysis |
135 |
|
|
Background |
135 |
|
|
Methods |
135 |
|
|
Dataset |
136 |
|
|
Results |
136 |
|
|
Discussion |
136 |
|
|
Summary |
137 |
|
|
Part III Applications: Emerging Opportunities |
139 |
|
|
Automatic Segmentation of Multiple Organs on 3D CT Images by Using Deep Learning Approaches |
140 |
|
|
Introduction |
141 |
|
|
Issue of Deep Learning for CT Image Segmentation |
141 |
|
|
Two Approaches for Multiple Organ Segmentations Using 2D and 3D Deep CNNs on CT Images |
142 |
|
|
Overview |
142 |
|
|
Deep Learning Anatomical Structures on 2D Sectional Images |
142 |
|
|
Deep Learning Local Appearances of Multiple Organs on 3D CT Images |
143 |
|
|
Conventional Image Segmentation Approach |
145 |
|
|
Results |
145 |
|
|
Discussions |
147 |
|
|
Segmentation Performances |
147 |
|
|
Training Protocol and Transfer Learning |
148 |
|
|
Comparison to Conventional Methods |
149 |
|
|
Computational Efficiency |
150 |
|
|
Conclusion |
151 |
|
|
Techniques and Applications in Skin OCT Analysis |
153 |
|
|
Introduction |
154 |
|
|
Skin Layer Segmentation in OCT |
154 |
|
|
Applications: Roughness, ET |
160 |
|
|
Deep Convolutional Networks in Skin Imaging |
161 |
|
|
Deep Learning for Classification of Dermoscopy Images |
162 |
|
|
Deep Learning for Classification of Full Field OCT Images |
162 |
|
|
Classification of Cross-Sectional OCT 2D Scans |
162 |
|
|
Semantic Segmentation in Cross-Sectional OCT Images |
164 |
|
|
Challenges |
164 |
|
|
Conclusions |
165 |
|
|
Deep Learning Technique for Musculoskeletal Analysis |
168 |
|
|
Importance of Musculoskeletal Analysis and Skeletal Muscle Analysis |
168 |
|
|
Musculoskeletal Recognition by Handcrafted Features and Its Limitations |
169 |
|
|
Skeletal Muscle Segmentation Using Deep Learning |
170 |
|
|
Whole-Body Muscle Analysis Using Deep Learning |
174 |
|
|
Fusion of Deep Learning and Handcrafted Features in Skeletal Muscle Modeling |
176 |
|
|
Conclusion |
177 |
|
|
Index |
180 |
|