Episode Details

In this lesson, Professor John Onofrey delves into the unique practical issues and challenges that Artificial Intelligence and Machine Learning applications encounter throughout the medical software life cycle. He contrasts these with traditional software development, highlighting the critical role of training data, risk assessment from sources like distributional shifts and adversarial attacks, and the necessity for robust system design to manage the inherent 'black box' nature and potential failures of AI/ML models in clinical settings.

Uploaded 22-08-25

🎯 Learning Objectives
• Differentiate the unique requirements and challenges of AI/ML applications within the medical software life cycle from traditional software.
• Identify and understand critical risk factors for AI/ML systems, such as distributional shifts, adversarial attacks, and performance drift.
• Explain the intrinsic link between training data, model development, and the importance of reproducibility and quality control.
• Articulate the concept of treating AI/ML components as "unsafe" and designing robust systems with failure mitigation strategies, including human-in-the-loop approaches.
• Describe the distinct workflows for training and deploying AI/ML models and the need for continuous post-market monitoring.



Learn more about your ad choices. Visit megaphone.fm/adchoices

Learn more about your ad choices. Visit megaphone.fm/adchoices