Foundation models in Radiology: Applications & Training

This lecture explores the transformative potential of foundation models (FMs) in radiology, contrasting them with task-specific AI. It outlines their evolution from CNNs to multimodal systems and demonstrates their ability to adapt across tasks such as segmentation, detection, and prediction of cancer outcomes. Multimodal FMs like RadFM and RadGPT are highlighted for generating detailed imaging reports and patient-friendly summaries, while agentic AI models exhibit diagnostic reasoning...