AI and machine learning can make diagnosis and treatment more effective and easier to access, but FDA regulation of AI/ML has not kept pace. FDA has regulated medical software through regulations and guidance for years, yet these were built for software that stays as validated.
FDA approves a final, validated version, while AI/ML is meant to learn and update after deployment, so the version in the field is no longer the one that was approved. This 60-minute webinar explains the current requirements, why they don't control AI/ML well, and the approaches FDA has been considering, which your development program should follow now.
Drawing on approved De Novo applications, it then shows how to get an AI/ML program approved and what submission documents you need. It is not a programming course.
WHY SHOULD YOU ATTEND?
It is not clear how to get an AI/ML program approved, because existing regulatory requirements don't adequately control software that keeps learning.
This session explains the approaches FDA has been considering and how to get your AI/ML program through FDA now, including the documentation a submission needs. It suits managers, directors and vice presidents in software engineering, engineering, regulatory affairs, quality assurance, marketing and management.
AREA COVERED
- Total product life cycle approach to AI/ML design
- Applying the FDA software Pre-Cert program to AI/ML
- The FDA discussion paper on AI/ML
- Database management
- QC of datasets
- Algorithm updating
- Reference standard development
- Standalone performance testing
- Clinical performance testing
- Data enrichment
- Emphasis on explainability
- Additional labeling requirements
- Cybersecurity
LEARNING OBJECTIVES
- Explain why current FDA software rules do not adequately control AI/ML
- Apply a total product life cycle approach to AI/ML design
- Describe the FDA Pre-Cert program and discussion paper concepts for AI/ML
- Manage databases, dataset QC, reference standards and algorithm updates
- Plan standalone and clinical performance testing
- Address explainability, labeling and cybersecurity requirements
- Prepare the documentation for an AI/ML submission
WHO WILL BENEFIT?
Managers, Supervisors, Directors, and Vice-Presidents in the areas of:
- Software Engineers
- Engineers
- Regulatory Personnel
- Quality Assurance Personnel
- Marketing
- Management
It is not clear how to get an AI/ML program approved, because existing regulatory requirements don't adequately control software that keeps learning.
This session explains the approaches FDA has been considering and how to get your AI/ML program through FDA now, including the documentation a submission needs. It suits managers, directors and vice presidents in software engineering, engineering, regulatory affairs, quality assurance, marketing and management.
- Total product life cycle approach to AI/ML design
- Applying the FDA software Pre-Cert program to AI/ML
- The FDA discussion paper on AI/ML
- Database management
- QC of datasets
- Algorithm updating
- Reference standard development
- Standalone performance testing
- Clinical performance testing
- Data enrichment
- Emphasis on explainability
- Additional labeling requirements
- Cybersecurity
- Explain why current FDA software rules do not adequately control AI/ML
- Apply a total product life cycle approach to AI/ML design
- Describe the FDA Pre-Cert program and discussion paper concepts for AI/ML
- Manage databases, dataset QC, reference standards and algorithm updates
- Plan standalone and clinical performance testing
- Address explainability, labeling and cybersecurity requirements
- Prepare the documentation for an AI/ML submission
Managers, Supervisors, Directors, and Vice-Presidents in the areas of:
- Software Engineers
- Engineers
- Regulatory Personnel
- Quality Assurance Personnel
- Marketing
- Management