Artificial Intelligence/Machine Learning in Nuclear Medicine and Hybrid Imaging: Using ANSI C and the Arduino Development Environment

Description

This book includes detailed explanations of the underlying technologies and concepts used in Artificial Intelligence (AI) and Machine Learning (ML) in the context of nuclear medicine and hybrid imaging. A diverse team of authors, including pioneers in the field and respected experts from leading international institutions, share their insights, opinions and outlooks on this exciting topic. A wide range of clinical applications are discussed, from brain applications to body indications, as well as the applicability of AI and ML for cardio-vascular conditions. The book also considers the potential impact of theranostics. To balance the technology-heavy and disease-specific applications, it also discusses ethical / legal issues, economic realities and the human factor, the physician. Though this discussion is not based on research and outcomes, it provides important insights into the ramifications of how AI and ML could transform Nuclear Medicine and Hybrid Imaging practice. As the first work highlighting the role of these concepts specifically in this field, rather than for medical imaging in general, this book offers a valuable resource for Nuclear Medicine Physicians, Radiologists, Physicists, Medical Imaging Administrators and Nuclear Medicine Technologists alike. Foreword Preface: Benefits and Challenges of AI/ML in Hybrid Imaging and Molecular Imaging Contents Part I: Technology 1: Role and Influence of Artificial Intelligence in Healthcare, Hybrid Imaging, and Molecular Imaging 1.1 AI Applications Support the Infrastructure and Interventions of Healthcare, Including Molecular Imaging 1.1.1 Drug Development 1.1.2 Clinical Workflow 1.2 AI’s Clinical Applications with a Focus on Molecular Imaging 1.2.1 Understanding Disease 1.2.2 Diagnosis 1.2.3 Radiologic-Pathology Correlation 1.2.4 Characterization 1.2.5 Treatment Planning 1.2.6 Prediction of Response to Treatment 1.2.7 Overall Prognosis 1.2.8 Reporting 1.3 Conclusion References 2: Introduction to Machine Learning: Definitions and Hybrid Imaging Applications 2.1 Introduction 2.2 History and Basic Definitions 2.3 Learning Paradigms 2.4 General Concepts of Machine Learning Methods 2.5 Classical Machine Learning Approaches 2.6 Artificial Neural Networks 2.7 Radiomics and Radiogenomics 2.8 Imaging Applications 2.9 Conclusions and Perspectives References 3: Radiomics in Nuclear Medicine, Robustness, Reproducibility, and Standardization 3.1 Introduction 3.2 Robustness of Radiomic Features 3.3 Image Acquisition 3.4 Image Reconstruction 3.5 Segmentation 3.6 Image Processing 3.7 Discretization 3.8 Software 3.9 Pitfalls 3.10 Standardization 3.11 Discussion 3.12 Conclusion References 4: Evolution of AI in Medical Imaging 4.1 Disease Characterization 4.2 Segmentation 4.3 Image Generation/Reconstruction 4.4 Data Corrections 4.5 Image Registration 4.6 Radiology Reporting 4.7 Conclusion References 5: The Basic Principles of Machine Learning 5.1 Introduction 5.1.1 The Task of ML 5.1.1.1 A Question 5.1.1.2 A Computer 5.1.1.3 An Algorithm or Model 5.1.1.4 Data to Interpret 5.1.2 Supervised Learning 5.1.3 Unsupervised Learning 5.1.4 Radiomics and Texture Analysis 5.1.5 Feature Reduction 5.1.6 Scaling and Normalization 5.1.7 Training, Validation, and Testing 5.2 Linear Regression 5.2.1 Under- and Overfitting 5.2.2 Linear Regression Mathematics 5.2.3 The Neural Network 5.2.4 The Objective Function 5.2.5 Gradient Descent 5.2.6 Deep Learning with Convolutional Neural Networks 5.2.7 Advanced Deep Learning Architectures 5.2.7.1 Autoencoders 5.2.7.2 ResNet 5.2.7.3 U-Net 5.2.7.4 Generative Adversarial Networks 5.2.7.5 Deep Boltzmann Machines 5.2.8 Deep Learning in Medical Image Analysis 5.2.8.1 Classification, Localization and Detection 5.2.8.2 Segmentation 5.2.8.3 Registration 5.2.8.4 Image Synthesis 5.2.9 Federated Learning References Part II: Clinical Applications 6: Imaging Biomarkers and Their Meaning for Molecular Imaging 6.1 Introduction 6.2 Imaging Biomarkers, Paradigm Shift in Medical Imaging 6.3 Imaging Biomarkers in Hybrid Molecular Imaging References 7: Integration of Artificial Intelligence, Machine Learning, and Deep Learning into Clinically Routine Molecular Imaging 7.1 Introduction 7.2 Classification 7.3 Segmentation 7.4 Detection and Localization 7.5 Applications of ML and DL in Molecular Imaging 7.6 Internal Department Applications 7.7 A Glance at Tomorrow 7.8 Workforce; Redundancy, Displacement, Transformation, and Opportunity 7.9 Summary References 8: Imaging Biobanks for Molecular Imaging: How to Integrate ML/AI into Our Databases 8.1 Introduction 8.2 Imaging Biobanks in Molecular Imaging 8.3 Bioethical Issues 8.4 Proposed Architecture References 9: Artificial Intelligence/Machine Learning in Nuclear Medicine 9.1 Introduction 9.2 Classification 9.2.1 Alzheimer’s Disease 9.2.2 Parkinson’s Disease 9.3 Segmentation 9.4 Image Generation and Processing 9.5 Low-Dose Imaging References 10: AI/ML Imaging Applications in Body Oncology 10.1 General Principles 10.2 Brain 10.2.1 Glioma 10.3 Neck 10.3.1 Head and Neck Cancer 10.3.2 Thyroid Cancer 10.4 Thorax 10.4.1 Lung Cancer 10.5 Abdomen 10.5.1 Esophageal Cancer 10.5.2 Liver Tumor 10.5.3 Prostate Cancer 10.6 Skeleton 10.6.1 Bone Metastases 10.7 Hematopoietic System 10.7.1 Lymphoma 10.7.2 Multiple Myeloma References 11: Artificial Intelligence/Machine Learning in Nuclear Medicine and Hybrid Imaging 11.1 Introduction to AI 11.2 AI to Improve Image Quality and Processing 11.2.1 Image Denoising 11.2.2 Image Reconstruction 11.2.3 AI Applications in Attenuation Correction 11.2.4 Image Segmentation 11.2.5 CT Segmentation: Coronary Artery Calcium 11.2.6 CT Segmentation: Epicardial Adipose Tissue 11.3 AI to Improve Physician Interpretation 11.3.1 Structured Reporting 11.3.2 Disease Diagnosis 11.3.3 Risk Prediction 11.4 Protocol Optimization: Application to Rest Scan Cancellation 11.5 Explainable AI 11.6 Summary References Part III: Impact of AI and ML on Molecular Imaging and Theranostics 12: Artificial Intelligence Will Improve Molecular Imaging, Therapy and Theranostics. Which Are the Biggest Advantages for Therapy? 12.1 Introduction 12.2 Literature Review 12.2.1 Morphological and Metabolic Tumor Volume Tracking 12.2.1.1 Volumetry-Based Oncological Response Assessment Frameworks 12.2.1.2 Automated Segmentation-Based Volumetry Techniques 12.2.1.3 Evolution of Automated Segmentation Using Neural Networks 12.3 Quantitative Image and Texture Analysis in Oncological Therapy Response Monitoring 12.3.1 Neuro-Oncology 12.3.2 Head and Neck Cancers 12.3.3 Lung Cancer 12.3.4 Prostate Cancer 12.3.5 Breast Cancer 12.3.6 Gastrointestinal Oncology 12.4 Discussion and Outlook References 13: Integrative Computational Biology, AI, and Radiomics: Building Explainable Models by Integration of Imaging, Omics, and Clinical Data 13.1 Introduction 13.2 Artificial Intelligence and Data-Driven Science 13.3 Multimodal Imaging and Radiomics 13.4 Integrative Computational Biology 13.5 Patient-Centric Medicine: Preventive and Data-Driven References 14: Legal and Ethical Aspects of Machine Learning: Who Owns the Data? 14.1 Introduction 14.2 Opening the “Ethics Bubble”: What Are the Concerns? 14.3 Going Beyond FAT: Beyond Medical Ethics 14.4 Who Owns Patient Data? 14.5 Conclusion References 15: Artificial Intelligence and the Nuclear Medicine Physician: Clever Is as Clever Does 15.1 I Am Looking Forward to More A.I. in My Practice Because… 15.1.1 The Images Will Look Prettier 15.1.2 My Life Will Be Easier 15.1.3 My Patients Will Be Better Off 15.2 I Am Wary of More A.I. Because… 15.2.1 I Don’t Understand It 15.2.2 I Don’t Trust It 15.2.3 I Don’t Want It 15.3 How to Proceed? Let’s Be Practical! References