PathVG introduces a novel benchmark for pathology visual grounding that enables precise region-level detection based on natural language descriptions.

• Addresses limitations of existing pathology AI approaches by combining flexibility of language with precise region detection
• Connects text descriptions with specific pathological regions for more targeted analysis
• Creates a foundation for more accurate AI-assisted pathology diagnostics
• Enables pathologists to query specific cellular features using natural language

This research advances Medical AI by allowing more precise communication between clinicians and AI systems, potentially improving diagnostic accuracy and efficiency in pathology workflows.

PathVG: A New Benchmark and Dataset for Pathology Visual Grounding