Masterclass: Market Authorisation of AI-enabled Medical Devices In-house Training

Introduction to Artificial Intelligence

Regulation as a Strategy: AI Legislation as a Global Power Play

• Reasons to regulate AI
• Brussels Effect, the first mover advantage
• Impact on SMEs
• Horizontal versus vertical legislative approaches
• Big Tech’s stakes
• Standardization, the new battlefield
• Beijing Effect

Regulatory requirements for AI-enabled medical devices

• International: Organization of Economic Cooperation and Development (OECD), World Health Organization (WHO), International Medical Device Regulators Forum (IMDRF)
• National or regional: Brazil, Canada, China, European Union, Japan, Singapore, South Korea, Taiwan, Thailand, United Kingdom, United States of America
• Notified body initiatives

Best Practices for Developing AI-enabled Medical Devices

• Putting healthcare legislative initiatives on AI into practice
• Lessons learned and common pitfalls from regulatory submissions of AI-based medical devices
• Describing your use case, application, algorithm, and impact on clinical care
• Data origin, acquisition, identity, access, collection, attributes, shifts over time, limitations, bias, processing, augmentation, relevance and independence, annotation, testing, reference standard, integrity and data protection
• Training, tuning, and testing
• Clinical evaluation
• Statistical analysis (generalizability, subgroup analysis, statistical power)
• Usability considerations, interpretability and explainability
• Post-market considerations, updates and complaints
• Diagnostic versus prognosis and prediction tools

State-of-the-art data governance and data management

EU Artificial Intelligence Act

• AI Act messages for the board
• Scope
• Provider obligations
• Deployer obligations
• GPAI model provider obligations
• Standards and guidelines
• Notified body obligations
• Authorized representative obligations
• AI Act governance
• Regulatory sandboxes and Testing and Experimentation Facilities
• Timelines

Implementing the EU Artificial Intelligence Act

• A reference checklist with activities, timelines, and considerations to help medical device manufacturers and their economic operators implement the AI Act

Publicly Funded Supporting Initiatives for the AI Ecosystem

• Data Labs – Collaborative spaces for experimentation with data and AI models, supporting iterative development and validation
• EU Data Intermediation Services – Trusted entities facilitating secure and compliant data sharing between data holders and users
• EU Data Altruism Organisations – Certified organisations that collect and share data voluntarily provided by individuals and companies for the common good
• Data Labs – Bring together and organize data from different sources for AI developers. Provide data-related services
• AI Testbeds –infrastructure for testing AI health solutions (TEF Health, Notified Bodies, Union AI Testing Support Structures)
• European Digital Innovation Hubs (EDIHs) – Regional centres offering technical support, testing, and training for digital transformation, including AI in healthcare
• AI Factories – Initiatives providing end-to-end AI development support, from prototyping to deployment, often with a sectoral focus
• Regulatory Sandboxes – Controlled environments that allow developers to test AI-enabled medical technologies under regulatory supervision
• Public – Private Partnerships: publicly funded programs to accelerate innovation, create innovative ecosystems, and advance regulatory science

AI Standardization landscape

Adaptive AI-based medical devices

• Examples
• Change within and outside predefined boundaries
• Regulatory science
  • Change considerations
  • Interplay with foundation models and their providers
  • Predetermined Change Control Plan
  • Medical Device Production System
  • Regulatory challenges and how to overcome them