Medical Imaging: Artificial Intelligence, Image Recognition, and Machine Learning Techniques

Description

The book discusses varied topics pertaining to advanced or up-to-date techniques in medical imaging using artificial intelligence (AI), image recognition (IR) and machine learning (ML) algorithms/techniques. Further, coverage includes analysis of chest radiographs (chest x-rays) via stacked generalization models, TB type detection using slice separation approach, brain tumor image segmentation via deep learning, mammogram mass separation, epileptic seizures, breast ultrasound images, knee joint x-ray images, bone fracture detection and labeling, and diabetic retinopathy. It also reviews 3D imaging in biomedical applications and pathological medical imaging. Cover Half Title Title Page Copyright Page Contents Preface Editors 1: A Novel Stacked Model Ensemble for Improved TB Detection in Chest Radiographs 1.1 Introduction 1.2 Materials and Methods 1.2.1 Data Collection and Preprocessing 1.2.2 Proposal 1—Feature Extraction Using Local/Global Feature Descriptors and Classification Using SVM 1.2.3 Proposal 2—Feature Extraction and Classification Using a Customized CNN 1.2.4 Proposal 3—Feature Extraction Using Pre-Trained CNNs and Classification Using SVM 1.2.5 Proposal 4—Constructing Stacked Model Ensembles 1.3 Results and Discussion 1.4 Conclusion and Future Work Acknowledgments Conflict of Interest References 2: The Role of Artificial Intelligence (AI) in Medical Imaging: General Radiologic and Urologic Applications 2.1 Introduction to Artificial Intelligence (AI) 2.1.1 Terminology 2.1.2 Practical Costs 2.2 Artificial Intelligence in Medicine 2.3 Artificial Intelligence in Radiology 2.3.1 Extrinsic Factors to Image Interpretation 2.3.2 Intrinsic Factors to Image Quality 2.3.2.1 Geometry 2.3.2.2 Contrast 2.3.2.3 Background 2.3.3 Specific Technical Example of AI in Medical Imaging 2.4 Urologic Applications 2.5 Benefits vs. Disadvantages 2.6 Future Considerations References 3: Early Detection of Epileptic Seizures Based on Scalp EEG Signals 3.1 Introduction 3.2 Electroencephalogram 3.3 EEG Signal Processing 3.3.1 EEG Data Preprocessing 3.3.2 Feature Extraction 3.3.3 SVM Implementation 3.3.4 Performance Metrics 3.4 Results and Discussion 3.4.1 Comparison with Studies Following Similar Performance Metrics 3.4.2 Overall Performance 3.4.3 Using Epoch-Specific Values Instead of Event-Specific Values 3.5 Conclusion References 4: Fractal Analysis in Histology Classification of Non-Small Cell Lung Cancer 4.1 Introduction 4.2 Methodology 4.2.1 Image Analysis 4.2.2 Computation of Fractal Dimension 4.2.3 Extraction of Radiomics Features 4.2.4 Classification 4.2.5 Results 4.3 Conclusion References 5: Multi-Feature-Based Classification of Osteoarthritis in Knee Joint X-Ray Images 5.1 Introduction 5.2 Causes of OA 5.3 Levels of Knee OA 5.4 Proposed Work 5.5 Literature Survey 5.6 The Proposed Methodology 5.6.1 Noise Removal and Image Enhancement 5.6.2 Curvature-Based Feature Extraction Method 5.6.3 Segmentation of Image 5.6.4 Boundary Extraction 5.6.5 Edge Curvature Computation 5.6.6 Classification 5.6.7 Results and Discussion 5.6.7.1 Results for Abnormal Images 5.6.7.2 Results for Normal Images 5.6.7.3 Classification Results 5.7 Texture Analysis-Based Feature Extraction Method 5.7.1 Segmentation 5.7.2 Locating the Center of the Synovial Cavity 5.7.3 Feature Extraction 5.7.4 Classification 5.7.5 Results and Discussion 5.7.5.1 Results for Abnormal Image 5.7.5.2 Comparison 5.7.5.3 Failure Analysis 5.8 Conclusion Acknowledgment References 6: Detection and Classification of Non-Proliferative Diabetic Retinopathy Lesions 6.1 Introduction 6.2 Methodology 6.2.1 Preprocessing 6.2.1.1 RGB Color Separation 6.2.1.2 Mask Separation 6.2.1.3 Image Enhancement 6.2.1.4 Histogram Equalization 6.2.2 Removal of Optic Disc from Fundus Images 6.3 Detection of Microaneurysms 6.4 Detection of Hemorrhages 6.5 Detection of EXs 6.6 Extraction of Retinal Blood Vessels 6.7 Experimental Work 6.7.1 Extraction of Mask 6.7.2 Removal of OD 6.7.3 Detection of MAs 6.7.4 Detection of EXs 6.7.5 Detection of Hemorrhages 6.7.6 Statistical Techniques on NPDR Lesions 6.7.6.1 Statistical Techniques on MAs 6.7.7 Statistical Techniques on EXs 6.7.8 Statistical Techniques on Hemorrhages 6.7.9 Statistical Techniques on Retinal Blood Vessels 6.7.10 Grading NPDR Lesions Using ANN 6.7.11 K-Means Clustering 6.7.12 Performance Measurement by Receiver Operating Characteristic Curve 6.8 Conclusion References 7: Segmentation and Analysis of CT Images for Bone Fracture Detection and Labeling 7.1 Introduction 7.2 Clinical Aspects 7.2.1 Anatomy of Long Bone 7.2.2 CT Imaging 7.2.3 Types of Fractures 7.3 Literature Survey 7.4 Proposed Methodology 7.4.1 Data Acquisition 7.4.2 Data Annotation 7.4.3 Unwanted Artifacts Removal 7.4.3.1 Histogram Stretching 7.4.4 Bone Region Extraction and Labeling 7.4.4.1 Seed Point Selection and Spreading 7.4.4.2 Threshold Value Definition 7.4.4.3 Unique Label Assignment 7.5 Results 7.5.1 Application to Real Patient-Specific Images 7.5.2 Clinical Ground Truth 7.5.3 Comparison with State-of-the-Art Methods 7.6 Conclusions Acknowledgment References 8: A Systematic Review of 3D Imaging in Biomedical Applications 8.1 Introduction 8.2 Volumetric Data 8.2.1 Data Acquisition 8.2.2 Volume Data 8.2.3 Grid Structures 8.2.4 Volume Visualization 8.2.5 Steps in Volume Visualization 8.2.5.1 Data Acquisition and Dimension Reconstruction 8.2.5.2 Data Preprocessing and Extraction 8.2.5.3 View Definition 8.3 Indirect Volume Rendering (Surface Fitting) 8.3.1 Opaque Cubes (Cuberilles) 8.3.2 Contour Tracing 8.3.3 Marching Cube 8.4 Direct Volume Rendering 8.4.1 Raycasting 8.4.2 Splatting 8.4.3 Shear-Warp 8.4.4 Maximum Intensity Projection 8.4.5 3D Texture Mapping Volume 8.5 Recent Advances in Volume Visualization 8.5.1 Advances in Hardware (GPU)-Based Volume Rendering 8.5.1.1 The Need for GPU 8.5.1.2 Accelerators on the GPU 8.5.2 Advances in TFs 8.5.2.1 Image-Centric TFs 8.5.2.2 Data-Centric TFs 8.5.3 Generative Adversarial Networks(GANs) 8.6 Tools and Libraries for Volume Visualization 8.7 Conclusion and Future Directions References 9: A Review on the Evolution of Comprehensive Information for Digital Sliding of Pathology and Medical Image Segmentation 9.1 Introduction 9.1.1 Hurdles Encountered in Digitization of Pathological Information 9.2 Pathology Origins towards Whole Slide Imaging (WSI) 9.3 Digitalization of Pathology Imaging 9.4 Computational Analysis of Pathological Imaging 9.4.1 Demand for Scale 9.5 Management Infrastructure 9.5.1 WSI Acquisition, Management, and Exchange 9.5.2 Pathology Analytical Imaging Infrastructure 9.6 Investigation in Computerized Pathology Medical Imaging 9.6.1 Nephropathy Glomerulosclerosis: Integrative Using Active Contour 9.6.2 Mapping Molecules in the Tumor Microenvironment Using k-Means Clustering 9.6.3 Challenges Transversely Occurred in the Process 9.7 Conclusion References 10: Pathological Medical Image Segmentation: A Quick Review Based on Parametric Techniques 10.1 Introduction 10.1.1 Role and Flow of Medical Image Segmentation 10.1.2 Challenges in Medical Image Modalities for Segmentation 10.1.3 Architecture of Medical Image Modalities 10.1.3.1 MRI 10.1.3.2 Electron Microscopy 10.1.3.3 Computed Tomography 10.1.3.4 US 10.2 Medical Image Segmentation Techniques 10.2.1 Thresholding 10.2.2 Region Growing 10.2.3 Bayesian Approach 10.2.3.1 Maximum a Posteriori (MAP) 10.2.3.2 Markov Random Field (MRF) 10.2.3.3 Maximum Likelihood (ML) 10.2.3.4 Expectation Maximization (EM) 10.2.4 Clustering 10.2.4.1 k-Means 10.2.4.2 Fuzzy C-Means 10.2.5 Deformable Methods 10.2.5.1 Parametric Deformable Models (Explicit) 10.2.5.2 Non-Parametric Models (Implicit) 10.2.6 Atlas-Guided Approaches 10.2.6.1 Atlas as Average Shape 10.2.6.2 Atlas as Individual Image 10.2.7 Edge-Based Approaches 10.2.8 Compression-Based Approaches 10.2.9 Other Techniques 10.3 Study and Conversation 10.4 Comparison of Medical Image Segmentation Methods with Experimental Analysis 10.5 Conclusions References Index