Authors:Yusuf Salcan (1 and 4), Simon Ging (1 and 2), Robin Schirrmeister (3), Philipp Arnold (3), Elmar Kotter (3), Behzad Bozorgtabar (2), Thomas Brox (1) ((1) Computer Vision Group, University of Freiburg, Germany, (2) Adaptive & Agentic AI (A3) Lab, Aarhus University, Denmark, (3) Department of Radiology, Medical Center -- University of Freiburg, Germany, (4) CRIION-AI Lab, Freiburg, Germany)
Abstract:We study how to train visually grounded vision-language models (VLMs) for radiology without manual spatial annotations. We introduce RefRad2D, a large-scale bilingual (German/English) dataset of 1.2M CT and MR image-text pairs derived from clinical practice, with task-specific VQA and spatial grounding subsets generated automatically via LLM-based curation and automated segmentation. Trained on this data, our model RadGrounder jointly performs report generation, visual question answering, and spatial grounding via bounding-box detection or segmentation. On external VQA benchmarks (Slake, VQA-RAD), RadGrounder achieves competitive results with specialized medical VLMs. Adding our clinical data to the training mixture improves open-ended VQA over fine-tuning on the downstream datasets alone, showing the transferability of our dataset. Crucially, adding grounding supervision does not degrade language quality, enabling spatially verifiable outputs at no cost to VQA performance.
From: Simon Ging [view email]
[v1]
Thu, 18 Jun 2026 16:55:26 UTC (1,328 KB)
此内容由惯性聚合(RSS阅读器)自动聚合整理，仅供阅读参考。 原文来自 — 版权归原作者所有。