Turning Medical Imaging AI Breakthroughs into Bedside Impact via Physician-in-the-Loop AI

Explore a comprehensive keynote speech by Dr. Arman Rahmim delivered at the MICCAI 2025 HECKTOR Challenge that critically examines the current state of medical imaging AI and proposes transformative approaches for real-world clinical implementation. Discover why medical imaging AI requires fundamental rethinking and redoing to achieve meaningful bedside impact, moving beyond traditional approaches that often oversimplify physician workflows and decision-making processes. Learn about the critical importance of incorporating non-imaging data into AI systems and understand the complex realities of what physicians actually do in clinical practice. Delve into the concept of Physician-in-the-loop AI and its counterpart, AI-in-the-loop Physician, exploring how these paradigms can create more effective human-AI collaboration in healthcare settings. Examine active learning and continual learning methodologies that enable AI systems to adapt and improve through ongoing physician interaction. Understand the essential requirement for designing clinical trials to demonstrate the added value of AI solutions before deployment. Investigate whether foundation models truly live up to their expectations in medical imaging applications and explore strategies for large dataset management, including data sharing and generation approaches. Discover platforms specifically designed for Physician-in-the-loop AI implementation and consider the philosophical question of whether AI should serve as an extension of human capabilities. Learn about the crucial role of implementation science in successfully translating AI breakthroughs from research laboratories to clinical practice, ensuring that technological advances result in tangible improvements in patient care and clinical outcomes.