Collaboration Between Clinicians and Vision-Language Models in Radiology Report Generation

Explore the intersection of artificial intelligence and radiology through this Stanford University MedAI Group Exchange Session lecture examining how vision-language models can collaborate with clinicians in generating radiology reports. Learn about cutting-edge research addressing the global radiologist shortage by developing specialist language models capable of producing X-ray reports that match or exceed radiologist-written reports in normal outpatient settings, as evaluated by expert radiologist panels. Discover how Google DeepMind's research team has created an assistive framework enabling effective collaboration between clinicians and AI systems for radiology report generation, particularly for cases containing errors or requiring expert oversight. Gain insights into the technical approaches for fine-tuning vision-language models specifically for medical imaging applications, the evaluation methodologies used to assess AI-generated reports against human expert standards, and the practical implications for clinical workflow integration. Understand the broader context of AI applications in biomedical domains including dermatology, clinical documentation reasoning, and diagnostic dialogue through the perspective of a leading researcher with extensive experience in machine learning, computational neuroscience, and biomedical AI applications. Participate in interactive discussion and Q&A exploring the challenges, opportunities, and future directions for human-AI collaboration in radiology and medical imaging interpretation.