Introduction to AI and Machine Learning in Medical Imaging, 1st Edition

Vision Message by Prof Christoph Wald
Foreword by Prof Minerva Becker
Foreword by Dr Sikandar Shaikh
Contributors
Preface
Acknowledgements
Abbreviations


1 Foundation of Artificial Intelligence and Machine Learning

1.1 Introduction to Artificial Intelligence (AI)
What? An overview of AI
When? A brief history
Why? The need for AI
Where? Applications
How? Pre-requisites for the success of an AI model


1.2 Machine Learning (ML) and Deep Learning (DL) Basics
Computer-aided systems
Basics of the hierarchy of AI
Machine learning versus deep learning algorithms
Ml techniques: Supervised, unsupervised, reinforcement learning
Machine learning architecture and various models
Introduction to neural networks
Transfer learning


2 Big Data

2.1 Data Preparation Pipeline
Introduction to data
Ground truth
Data preparation
Data de-identification
Data pre-processing and cleaning
Data curation
Data annotation
Data storage
Open source datasets


2.2 Data Handling
Data augmentation
Techniques for noise reduction
Data standardization


2.3 Medical Image Analysis
Key computer vision tasks
Steps of computer vision working feature extraction



3 Model Training and Validation with Potential Impediments

3.1 Pipeline for Training and Testing a Model
Data types, data splitting and model training/validation/testing
Internal versus external validation, generalizability
Evaluating artificial intelligence performance metrics in medical imaging: a comprehensive review of classification, segmentation and object detection metrics
Evaluation Metrics for Generative AI Models


3.2 Deployment Strategies for Artificial Intelligence and Machine Learning Models in Medical Imaging Within Clinical Environments
Foundational deployment strategies
Seamless integration into existing workflow
Navigating the challenges of deployment
Phases of testing AI software in medical imaging


4 Interfacing AI With Hospital Systems

4.1 Integration of AI With Hospital Infrastructure
Merging imaging data with hospital infrastructure and electronic health records
Implementation of AI in clinical practice
Interface with hospital operations for workflow optimization
Ethical, legal and regulatory considerations in integrating AI with hospitals
Implementation challenges and strategies


4.2 Integration of AI Algorithms into Existing Radiology Workflows
Before image acquisition
During image acquisition
After image acquisition
AI-assisted reporting tools
Natural language processing (NLP) in radiology reporting
AI orchestration
Augmented intelligence (AugI)
Obstacles In Integrating AI into Radiology Workflows


5 Stumbling Blocks and Remedial Measures

5.1 Bias in AI Models
Data bias
Annotation/labelling bias
Algorithmic bias
Deployment bias
Bias and variance
Overfitting and underfitting
Black box problems and explainable AI (XAI)
AI hallucinations
Shadow AI
Cybersecurity and data protection




5.2 Ethical and Legal Considerations
Regulatory guidelines and compliance
Ethical implications and mitigation strategies
Impediments to the adoption of AI


5.3 Risk Mitigation Strategies
Bias mitigation strategies


6 Emerging Trends and Future Applications

6.1 Generative AI
Introduction to generative AI
Generative adversarial networks (GANS)
Diffusion models
Transformer models
Speech-to-text (STT) models
Conversational AI
Multimodal AI
Natural language processing (NLP)
Prompt engineering
Retrieval-augmented generation (RAG)
Agentic AI
Guardrails


6.2 Emerging Trends
Precision medicine and precision radiology
Opportunistic screening and preventive radiology
Augmented intelligence (AugI)
Federated learning (FL)
Edge computing
Internet of medical things
Blockchain technology (BCT): a distributed ledger technology
Quantum computing with AI (QAI)
Radiomics
Connectomics
Theranostics
Robotics
Digital twins
Democratization of AI
Extended reality and AI
Artificial intelligence and sustainability in radiology
Artificial intelligence and teleradiology
Opportunities for research, teaching and collaboration in AI-driven medical imaging
Key determinants for the success of any AI model
Evolving role of clinicians in an era of AI



Appendices
Appendix 1: Common Myths and Misconceptions
Appendix 2: FDA (food and drug administration) Clearance vs Approval vs Grants
Appendix 3: Advanced Explainability Techniques
Appendix 4: Emerging Clinical Applications
Appendix 5: Subspecialty Deep-dive
Appendix 6: Causal AI and Medical Imaging
Appendix 7: Global and Policy Perspectives
Appendix 8: AI Pitfalls and Failure Scenarios in Medical Imaging
Appendix 9: Case-Based Narratives Integrating AI and Medical Imaging
Appendix 10: Ethics and Decision-Making Scenarios (as per WHO and ICMR Ethical Guidelines)
Appendix 11: Workflow Walkthroughs
Appendix 12: AI Maturity Model for Medical Institutions
Appendix 13: AI Reliability Pyramid in Medical Imaging
Appendix 14: AI Value Equation
Appendix 15: Medical Imaging Data Value Pyramid
Appendix 16: Goyal – Imaging Data Readiness Score (G-ImDaRES) for AI development
Appendix 17: Diachronic AI
Appendix 18: Hallucination Pointers
Appendix 19: The Liability Allocation Ladder of AI in Medical Imaging
Appendix 20: Counterfactual AI in Medical Imaging
Appendix 21: Physical AI
Appendix 22:AI in Ultrasound Probes
Appendix 23: Vision Language Action Models (VLAMSs)
Appendix 24: Key Skills a Clinician Must Develop Alongside AI

Use Cases
Use Cases Related to AI and Medical Imaging
Use Case 1: Artificial Intelligence in Acute Stroke and Head Injury
Use Case 2: Artificial Intelligence in Acute/Emergency Thoracic Imaging
Use Case 3: CT Abdomen Organ Measurements
Use Case 4: MRI Spine Measurements
Use Case 5: X-ray Fracture Detection
Use Case 6: LLMs in Radiology Workflow
Use Case 7: Predicting Diffusion-Weighted Brain MR Images from T2-Weighted Images Using Convolutional Neural Networks
Use Case 8: Integrating AI into Clinical Practice Through a Universal Orchestration Layer
Use Case 9: LLMs Powered Tools for Orchestration and Clinical Decision Support
Use Case 10: AI-Based Chest Screening
Use Case 11: AI-Based Lung Nodule Triaging for Screening of Early Lung Cancer
Use Case 12: AI-based Tools for Neurological Emergencies
Use Case 13: Hypothetical Examples of Case Studies (Conceptual Prototypes)
Use Case 14: Convolutional Neural Networks (CNNs) for Brain Haemorrhage Classification in CT Images

Multiple Choice Questions
Bibliography
Index


ONLINE RESOURCES
Flash Cards
AI Research Templates:

1. Practical AI Research Templates for Residents and Clinicians

2. Dataset Creation Checklist

3. Consent Form Template for AI Imaging Studies

4. Model Evaluation Sheet

5. Bias Assessment Framework

6. AI Study Feasibility Form

7. Imaging Annotation Protocol Template

8. AI Reporting Checklist (Radiologist Version)

9. Vendor Evaluation Checklist

10. AI Deployment Readiness Checklist