EndoVGGT: GNN-Enhanced Depth Estimation for Surgical 3D Reconstruction

TL;DR

EndoVGGT introduces a GNN-based framework with a dynamic graph attention module to enhance 3D reconstruction of deformable tissues in surgery, improving accuracy and generalization over prior methods.

Contribution

The paper presents EndoVGGT, a novel geometry-centric approach with DeGAT that captures long-range tissue correlations for robust, domain-agnostic surgical 3D reconstruction.

Findings

Increased PSNR by 24.6% over prior methods.

Achieved strong zero-shot generalization to unseen datasets.

Enforced global consistency in non-rigid tissue deformation recovery.

Abstract

Accurate 3D reconstruction of deformable soft tissues is essential for surgical robotic perception. However, low-texture surfaces, specular highlights, and instrument occlusions often fragment geometric continuity, posing a challenge for existing fixed-topology approaches. To address this, we propose EndoVGGT, a geometry-centric framework equipped with a Deformation-aware Graph Attention (DeGAT) module. Rather than using static spatial neighborhoods, DeGAT dynamically constructs feature-space semantic graphs to capture long-range correlations among coherent tissue regions. This enables robust propagation of structural cues across occlusions, enforcing global consistency and improving non-rigid deformation recovery. Extensive experiments on SCARED show that our method significantly improves fidelity, increasing PSNR by 24.6% and SSIM by 9.1% over prior state-of-the-art. Crucially, EndoVGGT exhibits strong zero-shot cross-dataset generalization to the unseen SCARED and EndoNeRF domains, confirming that DeGAT learns domain-agnostic geometric priors. These results highlight the efficacy of dynamic feature-space modeling for consistent surgical 3D reconstruction.