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.
@article{arxiv.2603.24577,
title = {EndoVGGT: GNN-Enhanced Depth Estimation for Surgical 3D Reconstruction},
author = {Falong Fan and Yi Xie and Arnis Lektauers and Bo Liu and Jerzy Rozenblit},
journal= {arXiv preprint arXiv:2603.24577},
year = {2026}
}
Comments
We withdraw this submission due to significant errors in the presentation and logical structure of the paper. We found that the current version does not accurately convey the research findings and requires a major overhaul of the manuscript's methodology description and results analysis