Machine Learning for Medical Applications : Computer Vision, Image Processing, Disease Detection.

Format:

Book

Author/Creator:

Rajamanickam, Ranjith.

Series:

Advanced Mechanical Engineering Series

Advanced Mechanical Engineering Series ; v.14/2

Language:

English

Subjects (All):

Machine learning.

Diagnostic imaging.

Physical Description:

1 online resource (582 pages)

Edition:

1st ed.

Place of Publication:

Berlin/Boston : Walter de Gruyter GmbH, 2025.

Summary:

Machine Learning for Medical Applications - Volume II delves into the intersection of artificial intelligence, computer vision, and healthcare, offering a comprehensive exploration of how machine learning is revolutionizing disease detection and diagnostics.

Contents:

Intro

Contents

List of contributors

Deep learning in computer vision

1 Introduction

2 Overview of deep learning

2.1 Definition and scope of deep learning

2.2 Historical development and milestones

2.3 Key concepts and terminology

2.4 Differences between deep learning and traditional machine learning

2.5 Advantages and limitations of deep learning

3 Evolution of deep learning in computer vision

3.1 Early neural networks and their impact

3.2 The breakthrough of CNNs

3.3 Key innovations in deep learning architectures (e.g., AlexNet and VGGNet)

3.4 Major competitions and benchmarks (e.g., ImageNet and COCO)

3.5 The role of hardware advancements in deep learning

4 Core components of deep learning models

4.1 Neural network layers: input, hidden, and output layers

4.2 Activation functions: ReLU, sigmoid, and tanh

4.3 Convolutional layers and operations

4.4 Pooling layers and their functions

4.5 Fully connected layers and their role

5 Training deep learning models for computer vision

5.1 Data preparation and augmentation techniques

5.1.1 Data preparation

5.1.2 Data augmentation

5.2 Loss functions and optimization algorithms

5.2.1 Optimization algorithms

5.3 Gradient descent and backpropagation

5.3.1 Gradient descent

5.3.2 Backpropagation

5.4 Regularization techniques to prevent overfitting

5.5 Hyperparameter tuning and model evaluation

5.5.1 Model evaluation

6 Impact of deep learning on computer vision applications

6.1 Improvements in image classification accuracy

6.2 Advances in object detection and localization

6.3 Enhanced performance in image segmentation tasks

6.4 Innovations in generative models for image synthesis

6.5 Applications in real-time computer vision systems

7 Conclusion

References.

Deep learning for medical image segmentation

2 Definition and scope of medical image segmentation

2.1 What is medical image segmentation?

2.2 Scope and applications

2.2.1 Clinical applications

2.2.2 Research applications

2.3 Importance in healthcare

2.4 Segmentation tasks and challenges

2.4.1 Challenges in segmentation

2.5 Current trends and innovations

3 Traditional segmentation techniques

3.1 Thresholding methods

3.1.1 Global thresholding

3.1.2 Adaptive thresholding

3.1.3 Multilevel thresholding

3.1.4 Challenges and limitations

3.2 Region-based methods

3.2.1 Area development

3.2.2 Region merging

3.2.3 Watershed segmentation

3.2.4 Region splitting and merging

3.2.5 Active contours (snakes)

3.3 Edge detection techniques

3.4 Clustering-based approaches

3.4.1 k-Means clustering

3.4.2 Gaussian mixture models (GMMs)

3.4.3 Clustering fuzzy c-means (FCM)

3.4.4 Mean shift clustering

3.4.5 Watershed algorithm

3.5 Model-based methods

3.5.1 Active contours and snakes

3.5.2 Level set methods

3.5.3 Deformable models

3.5.4 Statistical shape models

3.5.5 Template matching

4 Evolution toward deep learning-based methods

4.1 Limitations of traditional methods

4.1.1 Sensitivity to noise and artifacts

4.1.2 Reduced flexibility to deal with complex hierarchy

4.1.3 Time analysis and dimension of growth

4.2 Introduction to machine learning in segmentation

4.2.1 Historical background and original machine learning techniques

4.2.2 Segmentations by supervised learning and segmentations by unsupervised learning

4.2.3 Limitations of early machine learning models

4.2.4 Migration of problems to deep learning paradigms

4.2.5 Impact and advancements

4.3 Transition to deep learning

4.3.1 Details of neural networks.

4.3.2 Introduction of U-Net and its variants

4.3.3 Impact of GANs

4.3.4 Transfer learning and pretrained model integration

4.3.5 Challenges and opportunities in deep learning

4.4 Deep learning architectures for segmentation

4.4.1 Convolutional neural networks

4.4.2 U-Net architecture

4.4.3 Attention mechanisms

4.4.4 DeepLab series

4.4.5 Transformers for segmentation

4.5 Adoption in clinical practice

4.5.1 Education into clinical practice

4.5.2 Effects on correctness of diagnosis

4.5.3 Challenges in implementation

4.5.4 Regulatory and ethical of healthcare

4.5.5 Examples and testaments

4.5.6 Future prospects

5 Overview of imaging modalities and their characteristics

5.1 Magnetic resonance imaging (MRI)

5.1.1 Principles of MRI technology

5.1.2 Applications in soft tissue imaging

5.1.3 Imaging sequences and contrast mechanisms

5.1.4 Issues in MRI segmentation

5.1.5 Recent improvements and change

5.2 Computed tomography

5.2.1 Principles of CT imaging

5.2.2 Advantages of CT imaging

5.2.3 Segmentation in CT imaging

5.3 Ultrasound imaging

5.3.1 Principles of ultrasound imaging

5.3.2 Tables in clinical application

5.3.3 Advantages of ultrasound imaging

5.3.4 Challenges and limitations

5.3.5 Advances and innovations

5.4 X-ray and other modalities

5.4.1 Characteristics and applications

5.4.2 Computed tomography

5.4.2.1 Characteristics and applications

5.4.3 Ultrasound imaging

5.4.3.1 Characteristics and applications

5.4.4 Magnetic resonance imaging

5.4.4.1 Characteristics and applications

5.4.5 Positron emission tomography

5.4.5.1 Characteristics and applications

5.5 Multimodal imaging approaches

5.5.1 Synergy of complementary information

5.5.2 Methods of using multiple modalities

5.5.3 Challenges and considerations.

6 Importance of accurate segmentation in clinical practice

6.1 Improving diagnostic accuracy

6.1.1 Enhanced detection of pathological features

6.1.2 Refinement of surgical planning

6.1.3 Radiotherapy treatment assistance

6.1.4 Supporting intensive long-term observations

6.2 Facilitating treatment planning

6.2.1 Optimization of radiotherapy

6.2.2 Personalized treatment approaches

6.2.3 Interoperability with complex overview techniques

6.2.4 Supervision and modifying management

6.3 Monitoring disease progression

6.3.1 Monitoring changes in and trends of the disease

6.3.2 Evaluating treatment efficacy

6.3.3 Detecting disease recurrence

6.3.4 Enhancing personalized medicine

6.4 Reducing manual effort and subjectivity

6.4.1 Organizing work to minimize repetitive activities

6.4.2 Minimizing human error

6.4.3 Enhancing reproducibility and consistency

6.4.4 Supporting the quantitative computation

6.5 Potential for personalized medicine

6.5.1 Adapting interventions for the person

6.5.2 Role in precision oncology

6.5.3 Enhancing surgical planning

6.5.4 Observing disease development

6.5.5 Advancements in multiomics integration

6.5.6 Opportunities and threats

Reference

Deep learning for image segmentation

2 Early developments in CNNs for image segmentation

2.1 Introduction to CNNs in computer vision

2.2 The emergence of FCNs

2.3 SegNet

2.4 Advancements with conditional random fields (CRFs)

2.5 Segmentation with patch-based CNNs

2.5.1 Challenges in patch-based segmentation

2.5.2 Advantages of patch-based approaches

2.5.3 Comparisons with fully convolutional networks

2.5.4 Applications and developments

3 U-Net and variants for biomedical image segmentation

3.1 Introduction to U-Net architecture.

3.1.1 Contracting path: encoder

3.1.2 Expansive path: decoder

3.1.3 Skip connections and their importance

3.1.4 Applications and impact

3.2 Extensions and improvements of U-Net

3.2.1 U-Net++: nested U-Net architecture

3.2.2 Attention U-Net: incorporating attention mechanisms

3.2.3 Multiscale U-Net: handling multiresolution inputs

3.2.4 Recurrent U-Net: adding recurrent layers

3.2.5 U-Net3D: toward the extension of U-Net for volumetric data

3.3 Training U-Net models efficiently

3.3.1 Data augmentation techniques

3.3.2 Optimization of loss functions

3.3.3 Regularization and hyperparameter tuning

3.3.4 Transfer learning and pretrained models

3.3.5 Computational resources and parallelization

3.4 Applications of U-Net beyond biomedical imaging

3.4.1 Satellite as well as aerial image segmentation

3.4.2 Self-driving cars and road environment understanding

3.4.3 Industrial and manufacturing applications

3.4.4 Robotics and drone uses

3.5 Challenges in U-Net architectures

3.5.1 Time and space complexity and time requirements

3.5.2 Overfitting and generalization issues

3.5.3 Handling variability in input data

3.5.4 Loss of fine details and spatial resolution

3.5.5 Scalability and adaptability

4 DeepLab family of architectures

4.1 Introduction to DeepLab architecture

4.1.1 Atrous convolutions and their impact

4.1.2 DeepLab's spatial pyramid pooling module

4.1.3 Advantages of DeepLab architecture

4.1.4 Evolution of DeepLab models

4.2 DeepLabv2

4.2.1 ASPP

4.2.2 Advantages over previous models

4.2.3 Implementation details and performance

4.2.4 Challenges and limitations

4.3 DeepLabv3

4.3.1 ASPP

4.3.2 Improved feature extraction with ASPP

4.3.3 Comparison with previous DeepLab models

4.3.4 Applications and performance

4.4 DeepLabv3+.

4.4.1 Architecture and enhancements.

Notes:

Description based on publisher supplied metadata and other sources.

Part of the metadata in this record was created by AI, based on the text of the resource.

ISBN:

3-11-220519-7

OCLC:

1534402911