Decouple and Reason: Anatomically Guided Two-Stage Voxel-Level Grounding of Free-Text Findings in 3D Chest CT

2026-07-14

AI Summary

This work addresses the challenge of precisely aligning free-text descriptions with voxel-level lesions in 3D chest CT scans. To overcome the coupling bottleneck between local feature extraction and semantic understanding inherent in existing end-to-end approaches, the authors propose a decoupled two-stage framework. The first stage performs class-agnostic 3D lesion segmentation, followed by a cross-modal reasoning step that aligns textual descriptions with the segmented regions. The method further incorporates lobar anatomical priors and relative spatial coordinate encoding to enhance spatial disambiguation of local regions. Evaluated on the ReXGroundingCT benchmark, the approach achieves state-of-the-art voxel-level grounding performance, demonstrating the effectiveness of both the decoupled design and the anatomy-guided alignment mechanism.

Abstract

Automatic voxel-level grounding of free-text findings in 3D chest Computed Tomography (CT) is critical for clinical interpretability. However, this task remains highly challenging due to the intricate spatial complexity of large 3D volumes and the heterogeneity of free-text findings. Existing end-to-end approaches often struggle to simultaneously learn the localized feature representations required for accurate 3D segmentation and the complex semantic understanding needed for text alignment, leading to suboptimal grounding performance. To overcome this fundamental limitation, we propose a novel decoupled framework that disentangles the problem into two specialized stages: (1) class-agnostic lesion segmentation and (2) text-volume reasoning. This structural separation allows the model to first extract candidate sub-volumes by localizing potential abnormalities. Subsequently, intensive cross-modal reasoning is performed to align these localized sub-volumes with free-text medical findings. To resolve the spatial ambiguities inherent in local regions, the reasoning module is augmented with explicit anatomical guidance, utilizing relative spatial coordinates and lung lobe priors. Evaluated on the ReXGroundingCT benchmark, our method achieves state-of-the-art performance in overall grounding quality on the official leaderboard. These results demonstrate that decoupling detection from reasoning is a highly effective paradigm for handling the complexity of 3D medical visual grounding. Our code is publicly available at https://github.com/khuhm/DAGG.

Problem

• 3D chest CT
• voxel-level grounding
• free-text findings
• medical visual grounding
• clinical interpretability

Innovation

• decoupled framework
• voxel-level grounding
• anatomical guidance
• cross-modal reasoning
• 3D chest CT