"Diff-RRG: Longitudinal Disease-wise Patch Difference as Guidance for LLM-based Radiology Report Generation” introduces a novel framework designed to better capture disease progression in chest X-rays. By extracting fine-grained, disease-specific patch differences between current and prior images, and explicitly guiding large language models with progression information, Diff-RRG generates reports that are both linguistically natural and clinically accurate. Experiments on the Longitudinal-MIMIC dataset demonstrate state-of-the-art performance, and our visualizations highlight how the model grounds its descriptions in specific pathological regions.
I am a Master's student in the Department of Artificial Intelligence at Korea University, and my research focuses on multimodal learning and medical image analysis, with a particular interest in radiology report generation.
It is a great honor to present my work at MICCAI 2025. Sharing my research on such a prestigious stage not only validates the importance of our work but also provides an invaluable step forward in my growth as an AI researcher.
I am particularly interested in advances in multimodal large language models, medical imaging analysis, and explainable AI approaches that can bring research outcomes closer to clinical application.
I look forward to engaging in insightful discussions with fellow researchers, exchanging ideas across different areas of medical imaging and AI. I hope these interactions will spark new collaborations and provide fresh perspectives that can enrich my future research.