Update cookies preferences

Meta Learning With Medical Imaging and Health Informatics Applications [Paperback / softback]

2.00/5 (2 ratings by Goodreads)
Edited by (Assistant Professor, Department of Electrical and Computer Engineering Department, University of Houston, USA), Edited by (University of Maryland, College Park, MD, USA), Edited by (University of Michigan, Ann Arbor, MI, USA)
  • Format: Paperback / softback, 428 pages, height x width: 235x191 mm, weight: 870 g, 60 illustrations (20 in full color); Illustrations
  • Series: The MICCAI Society book Series
  • Pub. Date: 29-Sep-2022
  • Publisher: Academic Press Inc
  • ISBN-10: 0323998518
  • ISBN-13: 9780323998512
Other books in subject:
  • Paperback / softback
  • Price: 137,73 €
  • This book is not in stock. Book will arrive in about 2-4 weeks. Please allow another 2 weeks for shipping outside Estonia.
  • Quantity:
  • Add to basket
  • Delivery time 4-6 weeks
  • Add to Wishlist
Meta Learning With Medical Imaging and Health Informatics ApplicationsLarger Image
  • Format: Paperback / softback, 428 pages, height x width: 235x191 mm, weight: 870 g, 60 illustrations (20 in full color); Illustrations
  • Series: The MICCAI Society book Series
  • Pub. Date: 29-Sep-2022
  • Publisher: Academic Press Inc
  • ISBN-10: 0323998518
  • ISBN-13: 9780323998512
Other books in subject:
Permanent link: https://www.kriso.ee/db/9780323998512.html
Keywords:
Meta-Learning, or learning to learn, has become increasingly popular in recent years. Instead of building AI systems from scratch for each machine learning task, Meta-Learning constructs computational mechanisms to systematically and efficiently adapt to new tasks. The meta-learning paradigm has great potential to address deep neural networks’ fundamental challenges such as intensive data requirement, computationally expensive training, and limited capacity for transfer among tasks.

This book provides a concise summary of Meta-Learning theories and their diverse applications in medical imaging and health informatics. It covers the unifying theory of meta-learning and its popular variants such as model-agnostic learning, memory augmentation, prototypical networks, and learning to optimize. The book brings together thought leaders from both machine learning and health informatics fields to discuss the current state of Meta-Learning, its relevance to medical imaging and health informatics, and future directions. The book comes with a GitHub repository consisting of various code examples and documentation to help the audience to set up Meta-Learning algorithms for their applications quickly.
  • First book on applying Meta Learning to medical imaging
  • Pioneers in the field as contributing authors to explain the theory and its development
  • Has GitHub repository consisting of various code examples and documentation to help the audience to set up Meta-Learning algorithms for their applications quickly
Contributors xv
PART 1 INTRODUCTION TO META LEARNING
Chapter 1 Learning to learn in medical applications
3(24)
Azade Farshad
Yousef Yeganeh
Nassir Navab
1.1 Introduction
3(1)
1.2 Problem statement
3(2)
1.3 Background
5(4)
1.3.1 Metric learning
5(3)
1.3.2 Optimization-based learning
8(1)
1.3.3 Model-based learning
8(1)
1.4 Task construction in meta learning
9(1)
1.5 Representation learning in meta learning
10(1)
1.6 Unsupervised / self-supervised meta learning
11(1)
1.7 Meta learning applications
12(7)
1.7.1 Segmentation
12(4)
1.7.2 Few-shot image generation
16(1)
1.7.3 Other applications
16(3)
1.8 Relation to federated learning
19(1)
1.9 Discussion
20(1)
1.10 Conclusion and outlook
20(7)
References
20(7)
Chapter 2 Introduction to meta learning
27(10)
Pengyu Yuan
Hien Van Nguyen
2.1 History of meta learning
27(1)
2.2 Formal definition
28(2)
2.2.1 Supervised learning
28(1)
2.2.2 Meta learning
28(2)
2.3 Meta learning formulation
30(2)
2.3.1 Task distribution perspective
30(1)
2.3.2 Meta learning from bilevel optimization view
31(1)
2.3.3 Feed-forward model view
31(1)
2.4 Meta learning taxonomy
32(5)
References
33(4)
Chapter 3 Metric learning algorithms for meta learning
37(16)
Pengyu Yuan
Hien Van Nguyen
3.1 Siamese networks for meta learning
37(1)
3.1.1 Model architecture
37(1)
3.1.2 Training and testing procedure
38(1)
3.2 Matching networks
38(4)
3.2.1 Model architecture
38(2)
3.2.2 Full-context embeddings
40(1)
3.2.3 Episodic training strategy
41(1)
3.3 Prototypical networks
42(2)
3.3.1 Model architecture
42(1)
3.3.2 Reinterpretation as a linear model
43(1)
3.3.3 Comparison to matching networks
44(1)
3.4 Relational networks
44(2)
3.4.1 Model architecture
44(1)
3.4.2 Relation networks for meta learning
45(1)
3.4.3 Relationship to existing models
46(1)
3.5 Graph neural networks
46(7)
3.5.1 General setup
47(1)
3.5.2 Model architecture
47(2)
3.5.3 Graph neural networks for meta learning
49(1)
3.5.4 Relationship with existing models
49(2)
References
51(2)
Chapter 4 Meta learning by optimization
53(12)
Pengyu Yuan
Hien Van Nguyen
4.1 Optimization as model
53(2)
4.1.1 Model architecture
54(1)
4.1.2 Parameter sharing and training strategy
54(1)
4.2 Model-agnostic meta learning
55(3)
4.2.1 A model-agnostic meta learning algorithm
56(1)
4.2.2 Training strategy
57(1)
4.2.3 Prototypical MAML
58(1)
4.3 Almost no inner loop meta learning
58(7)
4.3.1 Freezing layer representations
59(1)
4.3.2 Representational similarity experiments
59(1)
4.3.3 Feature reuse happens early in learning
60(2)
References
62(3)
Chapter 5 Model-based meta learning
65(10)
Pengyu Yuan
Hien Van Nguyen
5.1 Memory-augmented neural networks
65(2)
5.1.1 Model architecture
65(1)
5.1.2 Least recently used access
66(1)
5.1.3 Training and testing procedure
66(1)
5.2 Dynamic few shot visual learning
67(4)
5.2.1 Problem setup
67(1)
5.2.2 Few-shot classification weight generator
68(2)
5.2.3 Training procedure
70(1)
5.3 Metanetworks
71(4)
5.3.1 Base learner
71(1)
5.3.2 Layer augmentation
72(1)
5.3.3 Metalearner
72(2)
References
74(1)
Chapter 6 Meta learning for domain generalization
75(14)
Swami Sankaranarayanan
Yogesh Balaji
6.1 Introduction
75(1)
6.2 Related work
76(1)
6.3 Problem setup
77(1)
6.4 Meta learning Domain Generalization (MLDG)
77(3)
6.4.1 First order interpretation
78(1)
6.4.2 Sequential extension
79(1)
6.5 Metaregularization
80(3)
6.5.1 Learning the regularizer
81(1)
6.5.2 Training the final model
82(1)
6.5.3 Summary of the training pipeline
83(1)
6.6 Experiments
83(2)
6.7 Conclusion
85(4)
References
86(3)
PART 2 META LEARNING FOR MEDICAL IMAGING
Chapter 7 Few-shot chest x-ray diagnosis using discriminative ensemble learning
89(28)
Angshuman Paul
Yu-Xing Tang
Thomas C. Shen
Ronald M. Summers
7.1 Introduction
89(1)
7.2 Related work
90(2)
7.2.1 Deep CNN-based chest x-ray diagnosis
90(1)
7.2.2 Few-shot learning
91(1)
7.3 Methods
92(9)
7.3.1 Coarse-learner
93(1)
7.3.2 Saliency-based classifier
94(4)
7.3.3 Few-shot learning
98(3)
7.4 Experiments & results
101(11)
7.4.1 Dataset
101(1)
7.4.2 Performance measures & comparisons
102(5)
7.4.3 Ablation studies
107(4)
7.4.4 Notes on clinical applications
111(1)
7.5 Conclusions
112(5)
Acknowledgment
112(1)
Appendix 7.A On feature selection for autoencoders
112(1)
References
113(4)
Chapter 8 Domain generalization of deep networks for medical image segmentation via meta learning
117(24)
Quande Liu
Qi Dou
Cheng Chen
Pheng-Ann Heng
8.1 Introduction
117(1)
8.2 Related work
118(1)
8.2.1 Domain generalization
118(1)
8.2.2 Federated learning
118(1)
8.3 Domain generalization with shape-aware meta learning
119(5)
8.3.1 Method
119(3)
8.3.2 Experimental results
122(2)
8.4 Federated domain generalization with meta learning in continuous frequency space
124(12)
8.4.1 Preliminary
124(1)
8.4.2 Method
125(5)
8.4.3 Experimental results
130(6)
8.5 Conclusions
136(1)
8.6 Acknowledgment
136(5)
References
137(4)
Chapter 9 Meta learning for adaptable lung nodule image analysis
141(20)
Aryan Mobiny
Hien Van Nguyen
9.1 Introduction
141(2)
9.2 Related work
143(1)
9.3 Methodology
144(5)
9.3.1 Memory-augmented capsule network
144(1)
9.3.2 FastCaps++ as feature extractor network
144(2)
9.3.3 Memory-augmented task network
146(2)
9.3.4 Episodic training with simulated domain shift
148(1)
9.4 Dataset
149(2)
9.4.1 Data-I: LUNA-16 lung nodule dataset
150(1)
9.4.2 Data-II: collected lung nodule dataset
150(1)
9.4.3 Data-III: collected incidental lung nodule dataset
151(1)
9.5 Experiments and results
151(4)
9.5.1 Training procedure and implementations details
151(2)
9.5.2 Baseline deep neural network performance
153(1)
9.5.3 Evaluation of adaptive classifier
153(2)
9.6 Discussion
155(2)
9.7 Conclusion
157(4)
References
157(4)
Chapter 10 Few-shot segmentation of 3D medical images
161(24)
Abhijit Guha Roy
Shayan Siddiqui
Sebastian Polsterl
Azade Farshad
Nassir Navab
Christian Wachinger
10.1 Introduction
161(2)
10.1.1 Background on few-shot segmentation
162(1)
10.1.2 Challenges for medical few-shot segmentation
162(1)
10.1.3 Contributions
163(1)
10.1.4 Overview
163(1)
10.2 Prior work
163(1)
10.2.1 Few-shot learning
163(1)
10.2.2 Few-shot segmentation using deep learning
164(1)
10.3 Method
164(6)
10.3.1 Problem setup for few-shot segmentation
165(1)
10.3.2 Architectural design
165(3)
10.3.3 Training strategy
168(1)
10.3.4 Volumetric segmentation strategy
169(1)
10.4 Dataset and experimental setup
170(1)
10.4.1 Dataset description
170(1)
10.4.2 Problem formulation
170(1)
10.4.3 Hyperparameters for training the network
171(1)
10.5 Experimental results and discussion
171(9)
10.5.1 `Squeeze & excitation' based interaction
171(2)
10.5.2 Effect of skip connections in the architecture
173(1)
10.5.3 Model complexity of the conditioner arm
174(1)
10.5.4 Effect of the number of support slice budget
174(1)
10.5.5 Comparison with existing approaches
175(2)
10.5.6 Comparison with upper bound model
177(1)
10.5.7 Qualitative results
178(1)
10.5.8 Dependence on support set
178(2)
10.5.9 Discussion on spatial SE as interaction blocks
180(1)
10.6 Conclusion
180(5)
List of IDs in the visceral dataset
181(1)
Support set
181(1)
Validation query set
181(1)
Test query set
181(1)
Acknowledgment
181(1)
References
181(4)
Chapter 11 Smart task design for meta learning medical image analysis systems
185(28)
Cuong C. Nguyen
Youssef Dawoud
Thanh-Toan Do
Jacinto C. Nascimento
Vasileios Belagiannis
Gustavo Carneiro
11.1 Introduction
185(1)
11.2 Literature review
186(3)
11.2.1 Meta learning
186(2)
11.2.2 Breast screening from DCE-MRI
188(1)
11.2.3 Microscopy image cell segmentation
189(1)
11.3 Background
189(2)
11.3.1 Task
189(1)
11.3.2 Meta learning
189(2)
11.4 Task-augmentation weakly-supervised meta learning
191(3)
11.5 Unsupervised task formation meta learning
194(1)
11.6 Cross-domain few-shot meta learning
195(3)
11.7 Experiments
198(7)
11.7.1 Datasets
198(1)
11.7.2 Implementation details
198(2)
11.7.3 Results
200(2)
11.7.4 Discussion
202(3)
11.8 Conclusions
205(8)
References
206(7)
PART 3 META LEARNING FOR BIOMEDICAL AND HEALTH INFORMATICS
Chapter 12 AGILE - a meta learning framework for few-shot brain cell classification
213(22)
Pengyu Yuan
Hien Van Nguyen
12.1 Introduction
213(1)
12.2 Related works
214(2)
12.2.1 Brain cell type classification
214(1)
12.2.2 Few-shot classification (FSC)
215(1)
12.2.3 Active learning
216(1)
12.3 Methodology
216(7)
12.3.1 Task-augmented meta learning
217(2)
12.3.2 Active learning with Bayesian uncertainty
219(4)
12.3.3 Binary cell type classifier
223(1)
12.4 Experiments and results
223(9)
12.4.1 Datasets
223(3)
12.4.2 Baselines and metrics
226(1)
12.4.3 Rat brain cell FSC
227(4)
12.4.4 Human brain cell FSC
231(1)
12.5 Conclusion
232(3)
References
232(3)
Chapter 13 Few-shot learning for dermatological disease diagnosis
235(18)
Viraj Prabhu
Anitha Kannan
Murali Ravuri
Manish Chablani
David Sontag
Xavier Amatriain
13.1 Introduction
235(2)
13.2 Related work
237(1)
13.3 Approach
238(3)
13.3.1 Model
238(2)
13.3.2 Understanding the role of multiple clusters
240(1)
13.4 Results
241(5)
13.4.1 Experimental setup
241(2)
13.4.2 Main results
243(1)
13.4.3 Comparison between PCN and PN
244(2)
13.5 Per-class accuracy
246(1)
13.6 Role of hyperparameters
246(4)
13.6.1 Qualitative results
247(3)
13.7 Conclusion
250(3)
References
251(2)
Chapter 14 Knowledge-guided meta learning for disease prediction
253(22)
Qiuling Suo
Hyun Jae Cho
Jingyuan Chou
Stefan Bekiranov
Chongzhi Zang
Aidong Zhang
14.1 Introduction
253(2)
14.2 Related work
255(1)
14.3 Analysis of meta learning on TCGA data
256(5)
14.3.1 Modified MAML for pan-cancer prediction
256(2)
14.3.2 Pan-cancer prediction
258(3)
14.4 Transfer learning vs. meta learning
261(2)
14.4.1 Experimental results
262(1)
14.5 Knowledge-guided meta learning for healthcare
263(7)
14.5.1 Methodology
264(3)
14.5.2 Experiments
267(3)
14.6 Conclusion
270(5)
References
270(5)
Chapter 15 Case study: few-shot pill recognition
275(26)
Andreas Pastor
Suiyi Ling
Jieum Kim
Patrick Le Callet
15.1 Introduction
275(2)
15.2 Related work
277(2)
15.3 The proposed model
279(8)
15.3.1 Pill segmentation and localization
279(2)
15.3.2 Multistream CNN for pill recognition
281(6)
15.4 Proposed CURE pill database
287(2)
15.5 Experimental results
289(6)
15.5.1 Pill segmentation
289(1)
15.5.2 Imprinted text detection & recognition
290(2)
15.5.3 Pill recognition
292(3)
15.6 Demonstration for few-shot pill recognition: the `Pill Finder' application
295(1)
15.7 Conclusion
296(5)
References
297(4)
Chapter 16 Meta learning for anomaly detection in fundus photographs
301(30)
Sarah Matta
Mathieu Lamard
Pierre-Henri Conze
Alexandre Le Guilcher
Vincent Ricquebourg
Anas-Alexis Benyoussef
Pascale Massin
Jean-Bernard Rottier
Beatrice Cochener
Gwenole Quellec
16.1 Introduction
301(1)
16.2 Related machine learning frameworks
302(1)
16.3 Screening networks
303(1)
16.3.1 The OPHDIAT screening network
304(1)
16.3.2 The OphtaMaine screening network
304(1)
16.4 Datasets
304(1)
16.4.1 The OPHDIAT dataset
304(1)
16.4.2 The OphtaMaine dataset
305(1)
16.5 From frequent to rare ocular anomaly detection
305(6)
16.5.1 Deep learning for frequent condition detection
307(1)
16.5.2 Feature space definition
308(1)
16.5.3 F-distributed stochastic neighbor embedding (t-SNE)
309(1)
16.5.4 Feature space dimension reduction
309(1)
16.5.5 Probability function estimation
310(1)
16.5.6 Detecting rare conditions in one image
310(1)
16.6 Experiments in the OPHDIAT dataset
311(6)
16.6.1 Reference, validation, and testing
311(1)
16.6.2 Parameter selection
312(1)
16.6.3 Detection performance
312(2)
16.6.4 Heatmap generation
314(1)
16.6.5 Comparison with other machine learning frameworks
315(2)
16.7 From specific to general ocular anomaly detection
317(2)
16.7.1 The anomaly detection algorithm
317(1)
16.7.2 Development of the anomaly detection algorithm using OPHDIAT dataset
317(1)
16.7.3 Performance comparison with the baseline method
318(1)
16.7.4 Adaptation of the anomaly detection algorithm to the general population
318(1)
16.7.5 Evaluation of deep learning algorithms
318(1)
16.8 Results
319(2)
16.9 Discussion
321(10)
References
326(5)
Chapter 17 Rare disease classification via difficulty-aware meta learning
331(20)
Xiaomeng Li
Lequan Yu
Yueming Jin
Chi-Wing Fu
Lei Xing
Pheng-Ann Heng
17.1 Introduction
331(2)
17.2 Related work
333(3)
17.2.1 Skin lesion classification and segmentation from dermoscopic images
333(1)
17.2.2 Few-shot learning on skin lesion classification
334(1)
17.2.3 Rare disease diagnosis
335(1)
17.2.4 Meta learning in medical image analysis
335(1)
17.3 Methodology
336(4)
17.3.1 Preliminary knowledges
336(1)
17.3.2 Datasets
337(1)
17.3.3 Difficulty-aware meta learning framework
337(2)
17.3.4 Metatraining details
339(1)
17.4 Experiments and results
340(3)
17.4.1 Case study 1: ISIC 2018 skin lesion dataset
340(2)
17.4.2 Case study 2: validation on real clinical data
342(1)
17.5 Discussions
343(1)
17.6 Conclusion
343(8)
Acknowledgments
343(1)
References
344(7)
PART 4 OTHER META LEARNING APPLICATIONS
Chapter 18 Improved MR image reconstruction using federated learning
351(18)
Pengfei Guo
Puyang Wang
Jinyuan Zhou
Shanshan Jiang
Vishal M. Pate
18.1 Introduction
351(2)
18.2 Related work
353(1)
18.3 Methodology
354(4)
18.3.1 FL-based MRI reconstruction
354(2)
18.3.2 FL-MR with cross-site modeling
356(1)
18.3.3 Training and implementation details
357(1)
18.4 Experiments and results
358(7)
18.4.1 Datasets
359(1)
18.4.2 Evaluation of the generalizability
360(2)
18.4.3 Evaluation of FL-based collaborations
362(1)
18.4.4 Ablation study
363(2)
18.5 Conclusion
365(4)
References
366(3)
Chapter 19 Neural architecture search for medical image applications
369(16)
Viet-Khoa Vo-Ho
Kashu Yamazaki
Hieu Hoang
Minh-Triet Tran
Ngan Le
19.1 Neural architecture search: background
369(4)
19.1.1 Search space
370(1)
19.1.2 Search strategy
371(1)
19.1.3 Evaluation strategy
372(1)
19.2 NAS for medical imaging
373(6)
19.2.1 NAS for medical image classification
374(1)
19.2.2 NAS for medical image segmentation
375(3)
19.2.3 NAS for other medical image applications
378(1)
19.3 Future perspectives
379(6)
Acknowledgment
381(1)
References
381(4)
Chapter 20 Meta learning in the big data regime
385(10)
Swami Sankaranarayanan
Yogesh Balaji
20.1 Introduction
385(1)
20.2 Metapseudo labels
385(3)
20.2.1 Approximating pseudo label gradient with meta learning
386(1)
20.2.2 Experimental setup
387(1)
20.2.3 Results and discussion
388(1)
20.3 Metabaseline
388(4)
20.3.1 Problem setup
389(1)
20.3.2 Classifier baseline
389(1)
20.3.3 Metabaseline approach
390(1)
20.3.4 Experimental setup
391(1)
20.3.5 Results and discussion
391(1)
20.4 Conclusion
392(3)
References
393(2)
Index 395
Dr. Hien Van Nguyen is an Assistant Professor of the Department of Electrical and Computer Engineering Department at the University of Houston. His research interests are at the intersection between machine learning, computer vision, and biomedical image analysis. He has published 45 peer-reviewed papers and received 12 U.S. patents. His research has received awards from the National Science Foundation and the National Institutes of Health. He has served as area chairs of MICCAI (2019, 2021) and organized a series of popular MICCAI tutorials including deep learning for medical imaging (2015), deep reinforcement learning for medical imaging (2018), Bayesian deep learning (2019). Dr. Summers received a BA in physics and the M.D. and Ph.D. degrees in medicine/anatomy and cell biology from the University of Pennsylvania. He completed a medical internship at the Presbyterian-University of Pennsylvania Hospital, Philadelphia, PA, a radiology residency at the University of Michigan, Ann Arbor, MI, and an MRI fellowship at Duke University. In 2000, he received the Presidential Early Career Award for Scientists and Engineers, presented by Dr. Neal Lane, President Clinton's science advisor. In 2012, he received the NIH Director's Award, presented by NIH Director Dr. Francis S. Collins. He is an editorial board member of the journals Radiology and Academic Radiology. He was a co-chair of the Computer-aided Diagnosis program of the annual SPIE Medical Imaging conference in 2010 and 2011. He has co-authored over 300 journal, review and conference proceedings articles, and is a co-inventor on 12 patents. Prof. Rama Chellappa received the B.E. (Hons.) degree from the University of Madras, India, in 1975 and the M.E. (Distinction) degree from Indian Institute of Science, Bangalore, in 1977. He received M.S.E.E. and Ph.D. Degrees in Electrical Engineering from Purdue University, West Lafayette, IN, in 1978 and 1981 respectively. Since 1991, he has been a Professor of Electrical Engineering and an affiliate Professor of Computer Science at University of Maryland, College Park. He is also affiliated with the Center for Automation Research (Director) and the Institute for Advanced Computer Studies (Permanent Member). In 2005, he was named a Minta Martin Professor of Engineering. Prior to joining the University of Maryland, he was an Assistant (1981-1986) and Associate Professor (1986-1991) and Director of the Signal and Image Processing Institute (1988-1990) at University of Southern California, Los Angeles. Over the last 29 years, he has published numerous book chapters, peer-reviewed journal and conference papers. He has co-authored and edited books on MRFs, face and gait recognition and collected works on image processing and analysis. His current research interests are face and gait analysis, markerless motion capture, 3D modeling from video, image and video-based recognition and exploitation and hyper spectral processing.