准的脑血管分割成为脑血管疾病诊治的重要辅助手段，受到研究者的广泛关注。深度学习是一种启发式方法，它鼓励研究人员通过驱动数据集从图像中得出答案。随着数据集和深度学习理论的不断发展，在脑血管分割方面取得了重要成果。为了全面分析新的脑血管分割，本书以深度学习为核心主题，涵盖了基于滑动窗口的模型、基于U-Net的模型、基于卷积经网络的其他模型、基于小样本数据集的模型、基于半监督或无监督学习的模型、基于征融合的模型、基于Transformer的模型和基于几何图形学的模型。本书组织了不同模型的发展，改进以及具体案例，探讨了领域的发展趋势和展望。
陈诚，博士后，硕士生导师，从事医学语义理解、多模态三维重建与应用等方面研究。主持国家自然科学基金、中国博士后科学基金、教育部重点实验室开放基金等项目。以第一作者身份在IEEE Trans. Med. Imaging、IEEE Trans. Ind. Inform.等发表SCI/CCF A类会议10余篇。
Chapter 1 Introduction for Cerebrovascular Segmentation 1
1.1 Overview 1
1.2 Background 2
1.3 Cerebrovascular Imaging Modalities 6
1.4 Open Source for Medical Images Segmentation 9
1.5 Discussion of Development Trend 12
1.6 Discussion of Quantitative Assessment 13
1.7 Challenges and Opportunities 16
1.8 Conclusions 17
Chapter 2 DL-based Cerebrovascular Segmentation Model 19
2.1 Sliding Window Based Models 19
2.2 U-Net Based Models 20
2.3 Other CNNs Based Models 24
2.4 Small-Sample Based Models 26
2.5 Semi-Supervised / Unsupervised Learning Models 28
2.6 Fusion Based Models 30
2.7 Transformer Based Models 31
2.8 Graphics Based Models 32
Chapter 3 Generative Consistency for Semi-Supervised Learning
Cerebrovascular Segmentation from TOF-MRA 35
3.1 Overview 35
3.2 Introduction 36
3.3 Methods 39
3.4 Experiments 47
3.5 Discussion 51
3.6 Conclusion 55
Chapter 4 A Learnable Gabor Convolution Kernel for Vessel
Segmentation 57
4.1 Overview 57
4.2 Introduction 58
4.3 Methods 60
4.4 Experiments and Discussion 69
4.5 Conclusion 79
Chapter 5 Integration-and Separation-Aware Adversarial Model
for Cerebrovascular Segmentation from TOF-MRA 80
5.1 Overview 80
5.2 Introduction 81
5.3 Methods 84
5.4 Datasets 90
5.5 Experiments and Results 90
5.6 Discussion 96
5.7 Conclusion 100
Chapter 6 Cerebrovascular Segmentation in Phase-Contrast
Magnetic Resonance Angiography by Multi-Feature
Fusion and Vessel Completion 101
6.1 Overview 101
6.2 Introduction 102
6.3 Methods 105
6.4 Results 113
6.5 Discussion 116
6.6 Conclusion 126
References 128
书单推荐
