New MedGrounder model advances generalised medical phrase grounding

Researchers have introduced MedGrounder, a novel model for generalised medical phrase grounding (GMPG). This new approach addresses limitations in existing systems by mapping sentences to zero, one, or multiple image regions, accommodating multi-region findings and non-groundable phrases like negations. MedGrounder was trained in a two-stage process, first on sentence-anatomy box alignment and then fine-tuned on sentence-human annotated box datasets. Experiments on PadChest-GR and MS-CXR datasets demonstrate its strong zero-shot transfer capabilities and superior performance over existing baselines, particularly for complex phrases, while requiring fewer human annotations. AI

IMPACT Enhances interpretability of radiological reports by improving the mapping of text to image regions.

RANK_REASON Academic paper introducing a new model and task formulation. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CL →

• Generalised Medical Phrase Grounding
• MedGrounder
• Medical Phrase Grounding
• MS-CXR
• PadChest-GR
• Referring Expression Comprehension
• Wenjun Zhang

• paper
• model release

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →