AI in Clinical Research

An introduction to artificial intelligence concepts, written for clinicians rather than engineers. This section answers the question: what is AI, and what can it do for my research?

• 1.1 What Is Artificial Intelligence? A Clinician's Primer
• 1.2 Supervised, Unsupervised, and Self-Supervised Learning
• 1.3 Common AI Architectures in Medicine
• 1.4 Foundation Models and Large Language Models in Clinical Research
• 1.5 Matching the Problem to the Approach
• 1.6 What AI Cannot Do: Limitations and Common Misconceptions

Building effective AI models begins with data. This section covers everything a clinical researcher needs to know about assembling, preparing, and safeguarding datasets.

• 2.1 Why Data Quality Matters More Than Model Complexity
• 2.2 How Much Data Do You Need?
• 2.3 Data Annotation and Labeling for Clinical AI
• 2.4 Class Imbalance: When Rare Conditions Dominate Your Research Question
• 2.5 Data Preprocessing and Augmentation
• 2.6 Multi-Site and Multi-Device Data: The Generalization Challenge
• 2.7 De-Identification, Privacy, and Regulatory Compliance for Research Data
• 2.8 Synthetic Data and Data Simulation

A practical guide to the model development lifecycle, written for clinical researchers who will collaborate with engineering teams rather than build models alone.

• 3.1 The AI Development Lifecycle: From Hypothesis to Deployment
• 3.2 Training, Validation, and Test Sets: Why You Need All Three
• 3.3 Transfer Learning: Standing on the Shoulders of Large Models
• 3.4 Hyperparameter Tuning and Model Selection
• 3.5 Overfitting, Underfitting, and Regularization
• 3.6 Working with an Engineering Team: The Clinician's Role
• 3.7 Tools and Infrastructure: What You Need to Know

Technical metrics for assessing how well an AI model performs its intended task. This section gives clinical researchers the vocabulary to critically evaluate published results and their own models.

• 4.1 Classification Metrics: Sensitivity, Specificity, and Beyond
• 4.2 The ROC Curve and AUC: What They Tell You and What They Hide
• 4.3 Segmentation Metrics: Dice, IoU, and Volumetric Measures
• 4.4 Calibration: Does a 90% Confidence Score Really Mean 90%?
• 4.5 Statistical Significance, Confidence Intervals, and Sample Size for AI Studies
• 4.6 Cross-Validation Strategies for Medical Data
• 4.7 Comparing Models: When Is One Model Truly Better Than Another?

Moving beyond technical performance to demonstrate that an AI model provides real clinical value. This section bridges the gap between model development and clinical evidence.

• 5.1 Technical Validation vs. Clinical Validation: Understanding the Difference
• 5.2 Designing Retrospective Validation Studies
• 5.3 Designing Prospective Clinical Validation Studies
• 5.4 Multi-Site External Validation
• 5.5 Reader Studies: Measuring AI's Impact on Clinical Decision-Making
• 5.6 Defining Clinically Meaningful Endpoints
• 5.7 Reference Standards and Ground Truth in Medical AI

Navigating the regulatory landscape for bringing an AI-based medical device to market. This section covers U.S. FDA requirements in depth, with orientation to international frameworks.

• 6.1 When Does Your AI Algorithm Become a Medical Device?
• 6.2 Device Classification: Class I, II, and III
• 6.3 The 510(k) Pathway: Substantial Equivalence to a Predicate
• 6.4 The De Novo Pathway: First-in-Class Devices
• 6.5 Premarket Approval (PMA) and When It Applies
• 6.6 Breakthrough Device Designation: Faster Review for Innovative AI
• 6.7 Good Machine Learning Practice (GMLP): The 10 Guiding Principles
• 6.8 Predetermined Change Control Plans (PCCP)
• 6.9 Software Lifecycle Requirements: IEC 62304
• 6.10 Design History File and Quality Management System Essentials
• 6.11 International Regulatory Frameworks: EU MDR, Health Canada, and Beyond
• 6.12 Post-Market Surveillance and Real-World Monitoring

Ensuring that AI systems are equitable, transparent, and trustworthy. This section addresses the scientific, ethical, and increasingly regulatory dimensions of responsible AI development.

• 7.1 Sources of Bias in Medical AI
• 7.2 Demographic Fairness: Performance Across Populations
• 7.3 Generalization Across Devices, Sites, and Geographies
• 7.4 Explainability and Interpretability in Clinical AI
• 7.5 Transparency and Reproducibility
• 7.6 Ethical Frameworks for AI in Healthcare

Getting an AI tool from the lab into clinical practice. This section covers the practical, technical, and human factors that determine whether a validated model actually improves care.

• 8.1 From Model to Product: The Implementation Gap
• 8.2 Clinical Workflow Analysis: Where Does AI Fit?
• 8.3 Interoperability: DICOM, HL7 FHIR, and Integration Standards
• 8.4 Human-AI Interaction and User Interface Design
• 8.5 Clinician Trust, Adoption, and Change Management
• 8.6 Monitoring AI Performance in Production

Standards and best practices for communicating AI research to the clinical community. This section helps researchers produce publications that are rigorous, reproducible, and useful.

• 9.1 Reporting Standards: CONSORT-AI, SPIRIT-AI, TRIPOD+AI, and STARD
• 9.2 Writing the Methods Section for an AI Study
• 9.3 Common Methodological Pitfalls in Medical AI Papers
• 9.4 Sharing Code, Data, and Models
• 9.5 Peer Review of AI Manuscripts: What Reviewers Look For

• A Glossary of AI and Machine Learning Terms for Clinicians
• B Recommended Reading and Courses
• C Regulatory Standards Quick Reference
• D Checklist: Is Your AI Study Ready for Publication?