About the Session
Developing AI and ML imaging solutions for clinical use requires careful planning, rigorous data practices, and a clear understanding of regulatory pathways. Many imaging informatics professionals, radiologists, and engineers have strong prototypes but struggle to navigate the transition to an FDA-cleared Software as a Medical Device (SaMD). Early decisions in dataset curation, annotation, documentation, and bias mitigation can determine whether a product advances smoothly or encounters costly delays.
This session provides a practical, end-to-end framework for bringing AI/ML imaging tools from concept to clinical adoption. Presenters will outline best practices for data selection and annotation, highlight key regulatory considerations for both 510(k) and De Novo pathways, and discuss strategies for designing validation studies that meet FDA expectations. Case-based lessons from cleared AI/ML devices will illustrate what works in real-world development, deployment, and post-market monitoring.
Objectives
- Identify best practices in data selection, annotation, and bias mitigation to support AI/ML SaMD development.
- Differentiate between FDA’s 510(k) and De Novo pathways and describe when each is appropriate.
- Explain common pitfalls in AI/ML SaMD submissions and strategies to prevent delays.
- Design a high-level roadmap from prototype to FDA submission, including validation studies.
- Explain how PCCPs and real-world evidence shape the lifecycle of AI/ML medical devices.
Presented By
