Cutting-Edge Computational Intelligence in Healthcare with Convolution and Kronecker Convolution-based Approaches.

Format:

Book

Author/Creator:

Prakash BTech, MTech.

Contributor:

Prakash, Allam Jaya; Plawiak, Pawel; Patro, Kiran Kumar

Language:

English

Subjects (All):

Deep learning (Machine learning).

Artificial intelligence.

Physical Description:

1 online resource (322 pages)

Edition:

1st ed.

Place of Publication:

Chantilly : Elsevier Science & Technology, 2026.

Summary:

Cutting-edge Computational Intelligence in Healthcare with Convolution and Kronecker Convolution-based Approaches focuses on the use of deep learning techniques in the field of medical imagine analysis.

Contents:

Front Cover

Cutting-Edge Computational Intelligence in Healthcare With Convolution and Kronecker Convolution-Based Approaches

Copyright Page

Dedication

Contents

List of contributors

Foreword

Preface

Acknowledgments

1 Foundational concepts

1 Introduction to deep learning in medical imaging

1.1 Introduction

1.1.1 Modalities of medical imaging

1.1.2 Deep learning in medical imaging

1.2 Taxonomy of deep learning in medical image analysis

1.2.1 Current trends of deep learning in medical imaging

1.2.2 Common evaluation metrics in medical imaging tasks

1.3 Deep learning algorithms in medical imaging

1.4 Case study

1.4.1 Test case

1.4.1.1 Objective: detection of physical fatigue using facial thermal images

1.4.1.1.1 Description

1.4.1.1.2 Results

1.5 Challenges and limitations

1.5.1 Imaging challenges

1.5.2 Deep learning challenges

1.5.3 Current limitations of deep learning in medical imaging

1.6 Future directions

1.7 Summary

1.8 Conclusions

References

2 Fundamentals of convolutional neural networks

2.1 Introduction

2.2 Basic principles of convolutional neural networks

2.2.1 Evolution from perceptron

2.2.2 Components of convolutional neural network

2.2.2.1 Convolutional layers

2.2.2.2 Inception layers

2.2.2.3 Pooling layers

2.2.2.4 Fully connected layers

2.2.2.5 Activation layers

2.3 Training convolutional neural networks

2.3.1 Backpropagation

2.3.2 Loss functions

2.3.2.1 Cross-entropy loss

2.3.2.2 Sparse categorical cross-entropy loss

2.3.2.3 Focal loss

2.3.3 Hyperparameters

2.3.3.1 Optimizers

2.3.3.1.1 Stochastic gradient descent

2.3.3.1.2 Root mean square propagation

2.3.3.1.3 Adaptive moment estimation

2.3.3.2 Batch size

2.3.3.3 Epoch

2.3.3.4 Validation split

2.3.3.5 Learning rate.

2.4 Overfitting mitigation strategies

2.4.1 Residual layers

2.4.2 Batch normalization

2.4.3 Dropout layers

2.4.4 L2 Regularization

2.4.5 Data augmentation

2.5 Case study-fundus image categorization

2.5.1 Dataset description

2.5.2 Challenges in classification

2.5.2.1 Variability in eye images

2.5.2.2 Noise

2.5.2.3 Data availability

2.5.2.4 Fine-grained classification requirements

2.5.2.5 Loss function

2.5.2.6 Real-time feasibility

2.5.2.7 Generalization

2.5.3 Transfer learning

2.5.4 Fine-tuning

2.6 Results

2.6.1 Comparative analysis

2.6.1.1 Accuracy

2.6.1.2 Precision

2.6.1.3 Recall

2.6.1.4 F1-score

2.6.1.5 Speed

2.6.1.6 Model size

2.6.2 Results with the best-performing networking

2.6.2.1 Confusion matrix

2.6.2.2 Comparative analysis for different choices of optimizer

2.6.2.3 Comparative analysis for different loss functions

2.6.2.4 Comparative analysis for different batch size

2.7 Conclusion

2 Advanced techniques in deep learning with kronecker convolutions

3 Kronecker convolutions ensemble vision transformer and 3D kronecker U-net for volumetric segmentation of kidney stones, cysts and tumor from CT scans

3.1 Introduction

3.2 Literature survey

3.2.1 Literature on kidney tumor segmentation

3.2.2 Literature on kidney stone segmentation

3.3 Methodology

3.3.1 Overall architecture

3.3.1.1 Scope reduction

3.4 Kronecker U-Net

3.4.1 Encoding

3.4.2 Skip connection

3.4.3 Bottleneck

3.4.4 Decoding

3.5 Experimental results

3.6 Experimental setup

3.7 Datasets

3.8 Preprocessing

3.9 Comparison of classification algorithms of all four datasets

3.10 State-of-the-art of kidney stone dataset

3.11 SOTA of KiTS19

3.12 SOTA of KiTS21 and KiTS23

3.13 Conclusion and future works

Acknowledgments.

References

4 Image processing techniques in healthcare for early detection of heart diseases

4.1 Introduction

4.2 Overview of heart diseases

4.3 Importance of early detection for effective treatment and prevention

4.3.1 Role of image processing in healthcare

4.3.2 The advantages of using image processing techniques for heart disease detection

4.4 Objectives of the chapter

4.5 Overview of medical imaging for heart disease

4.5.1 Types of imaging modalities

4.5.1.1 Electrocardiogram

4.5.1.2 Echocardiography

4.5.1.3 Computed tomography

4.5.1.4 Magnetic resonance imaging

4.5.1.5 Angiography

4.6 Challenges in medical imaging for cardiac diseases

4.6.1 Image processing techniques for early detection of heart diseases

4.6.1.1 Capsule networks for improved feature representation

4.6.1.2 Generative adversarial networks for data augmentation

4.6.1.3 Preprocessing in medical imaging for cardiac diseases

4.6.1.4 Deep learning-based methods

4.6.1.5 Enhancement

4.6.1.5.1 Contrast-limited adaptive histogram equalization

4.6.1.6 Feature extraction in medical imaging for cardiac diseases

4.6.1.7 Deep learning methods in cardiac disease detection

4.6.1.8 Denoising and restoration in cardiac imaging

4.6.1.8.1 Wavelet transforms

4.6.1.8.2 Total variation minimization

4.7 Applications in early detection of heart diseases

4.8 Datasets used for coronary artery disease, heart failure, and arrhythmia detection

4.8.1 Techniques used in cardiovascular image processing and their impact on model robustness

4.9 Future directions in cardiac imaging

4.10 Conclusion

3 Applications in medical imaging

5 Automated atypical teratoid /rhabdoid tumor detection in magnetic resonance imaging usingdeep learning

5.1 Introduction

5.2 Related work

5.3 Proposed methodology.

5.3.1 Data collection

5.3.2 Preprocessing

5.3.3 Model development

5.3.4 Training and optimizing model

5.3.5 Interpretability and integration

5.4 Result analysis

5.4.1 Evaluation metrics

5.5 Conclusion

6 Ischemic stroke lesion segmentation using multiscale processing and knowledge distillation through intra-domain teacher

6.1 Introduction

6.2 Related works

6.3 Methodology

6.3.1 Overall architecture

6.3.2 Multiscale block

6.3.3 Knowledge distillation-based intra-domain transfer learning

6.3.4 Loss functions

6.4 Datasets, metrics, and experimental configuration

6.4.1 Datasets

6.4.2 Evaluation metrics

6.4.3 Experimental setup

6.5 Results and discussion

6.5.1 Optimal hyperparameters

6.5.2 Quantitative results

6.5.3 Qualitative results

6.5.4 Ablation studies

6.5.5 P-R curve analysis

6.5.6 Computation time

6.5.7 Discussion

6.6 Conclusion

7 Disease classification through advanced neural networks

7.1 Introduction

7.1.1 Overview of ocular diseases

7.1.1.1 Age-related macular degeneration

7.1.1.2 Glaucoma

7.1.1.3 Optic atrophy

7.1.1.4 Diabetic retinopathy

7.1.2 Fundus image diagnostics

7.2 Advanced neural network architectures

7.3 GoogleNet

7.3.1 Inception V2

7.3.2 Inception V3

7.4 Residual networks

7.4.1 ResNet-50

7.4.2 ResNet-101

7.5 DenseNet

7.5.1 DenseNet-121

7.5.2 DenseNet-169

7.6 Xception

7.6.1 Xception block

7.6.2 Residual connections

7.7 ConvNeXT

7.8 Data augmentation

7.8.1 Rotation and flipping

7.8.2 Color jittering and contrast adjustments

7.9 Results

7.9.1 Real-time feasibility in clinical settings

7.10 Conclusion

4 Real-world implementation

8 GAT-Net: ghost attention network for classification of gait-based neurodegenerative diseases.

8.1 Introduction

8.2 Related works

8.3 Datasets

8.3.1 INIT Gait dataset

8.3.2 DAI Gait dataset

8.3.3 DAI2 Gait dataset

8.4 Methodology

8.4.1 Overview of network architecture

8.4.2 GAT block integration

8.4.3 Classification

8.5 Experiments and results

8.5.1 Performance evaluation on INIT dataset

8.5.2 Performance evaluation on DAI dataset

8.5.3 Performance evaluation on DAI2 dataset

8.5.4 Ablation

8.5.5 Performance comparison with state-of-the-art methods

8.6 Conclusion

9 Artificial intelligence-enhanced diagnostics: deep learning in medical imaging

9.1 Introduction

9.2 Deep learning model, characteristics, and interdependencies

9.3 Deep learning models

9.3.1 Convolutional neural networks

9.3.2 Recurrent neural networks

9.3.3 Attention models

9.3.4 Transformer structures

9.4 Training and architectural design

9.4.1 Evaluation metrics

9.4.2 Dataset

9.4.3 Deep learning frameworks

9.5 Applications of deep learning in medical imaging

9.5.1 Image classification

9.5.2 Image segmentation

9.5.3 Image registration

9.5.4 Image localization/detection

9.6 Challenges and future prospects

9.7 Conclusion

10 Precision medicine through imaging analytics: Kronecker convolutions in tumor detection

10.1 Introduction

10.1.1 Convolutional neural networks in healthcare

10.1.2 Synopsis of CNN architecture

10.1.3 Medical imaging applications

10.1.4 Fracture and lesion identification

10.1.5 Predictive analytics using convolutional neural networks

10.2 Mathematical demonstration of CNN's image classification efficacy

10.2.1 Local connectivity

10.2.2 Weight sharing

10.3 Kronecker convolution: a novel methodology

10.3.1 Mechanism of Kronecker convolution.

10.3.1.1 Advancement of traditional convolution.

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:

0-443-33083-2

0-443-33082-4

9780443330834

OCLC:

1573522431