FDA Recommendations for Artificial Intelligence/Machine Learning (AI/ML)-Enabled Device Software Functions

AI is expected to change healthcare, but it brings risks around ethical data use, biased algorithms, patient safety and cybersecurity. FDA has started overseeing AI/ML-enabled device software functions, and its recommendations center on a predetermined change control plan (PCCP) in the marketing submission.

This 75-minute webinar explains what a PCCP is and what it should contain: the types of modifications to include, the modification protocol, traceability and impact assessment for changes to machine learning models and algorithms.

You'll also look at data management and retraining practices, performance evaluation protocols and update procedures. Device examples and hypothetical scenarios show when a model change stays within an authorized PCCP and when a new marketing submission is needed.

• Explain why FDA expects a PCCP in marketing submissions for AI/ML device software
• Identify the types of modifications that belong in a PCCP
• Describe the contents of a PCCP, including the modification protocol and impact assessment
• Plan data management and retraining practices for ML device software functions
• Set up performance evaluation protocols and update procedures
• Tell when a model change is covered by an authorized PCCP and when a new submission is needed

• Machine learning, AI and AI/ML device software functions
• The AI Bill of Rights and design and development considerations for AI/ML software
• Marketing submission considerations for an FDA-recommended predetermined change control plan (PCCP)
• Types of modifications to include in a PCCP
• PCCP contents
• Modification protocol, traceability and impact assessment for changes to ML models and algorithms
• Data management, retraining, performance evaluation protocols and update procedures
• Device examples and hypothetical scenarios: compliant modifications vs a new marketing submission

A growing number of AI/ML-enabled devices, many in radiology, have FDA authorization. Manufacturers that anticipate post-approval changes in an authorized PCCP don't need a new submission for those changes, which saves time and cost.

The session suits regulatory affairs professionals, engineers, developers, executives and founders at AI and medical device companies, as well as researchers and students who want to understand FDA's expectations and the AI Bill of Rights.

• AI software companies in the medical device industry – executives, entrepreneurs, engineers, developers, programmers, research & development personnel
• Medical device companies – executives, entrepreneurs, engineers, developers, programmers, research & development personnel
• Regulatory Affairs Professionals in medical device and AI companies 
• Software engineers, biomedical engineers, software developers, programmers,
• IT personnel of AI machine learning software
• AI researchers and students at universities
• Students in regulatory programs

Rachelle D’Souza began her career in regulatory compliance in 2007. She has provided Regulatory Affairs, Quality Assurance, Clinical, Pharmacovigilance and Medical Information support to the Pharmaceutical, Biologic, Generic, Natural Health Product, Dietary Supplement, Medical Device, Cosmetic, Pesticide and Food industries.

Rachelle has developed regulatory strategies, independently prepared regulatory submissions and secured product and site approvals from global regulatory agencies. She has also designed, implemented, and maintained global quality, pharmacovigilance and medical information systems. Rachelle’s regulatory articles and webinars on the latest global

regulatory developments have been published in print and online by regulatory professional associations, webinar hosting platforms and industry magazines.