GLeVE: Graph-Guided Lesion Grounding with Proposal Verification in 3D CT

TL;DR

GLeVE is a novel framework that improves 3D CT lesion grounding by combining graph reasoning, anatomical verification, and octree refinement, leading to better localization and segmentation accuracy.

Contribution

It introduces a graph-guided, anatomy-aware lesion grounding method with hierarchical refinement, advancing beyond existing phrase-level and dense supervision approaches.

Findings

GLeVE outperforms classical multimodal models in lesion localization.

The method achieves higher segmentation accuracy on AbdomenAtlas 3.0.

Hierarchical octree refinement improves boundary delineation.

Abstract

Grounding radiology report descriptions to 3D CT volumes is essential for verifiable clinical interpretation, yet remains challenging due to the semantic-spatial gap between free-text narratives and volumetric anatomy. Existing report-assisted and vision-language grounding methods typically rely on phrase-level alignment or dense pixel supervision, resulting in limited lesion-wise correspondence and suboptimal localization accuracy. We propose GLeVE, a graph-guided lesion grounding framework with anatomical prior verification and octree-based autoregressive refinement. GLeVE treats each lesion description as an atomic semantic unit and encodes organ attribution, attributes, and inter-lesion relations through relation-aware graph reasoning to produce discriminative lesion-wise queries. Anatomy-aware proposal generation with region-level verification enforces one-to-one text-lesion alignment, while hierarchical octree refinement progressively improves boundary delineation. Experiments on AbdomenAtlas 3.0 demonstrate consistent gains over classical multimodal foundation models and report-supervised baselines in both segmentation accuracy and lesion-level localization.