Abstract
Dense Simultaneous Localization and Mapping (SLAM) is pivotal for providing 3D intraoperative guidance in endoscopic surgery. However, the unique challenges of the endoscopic environment characterized by weak textures, specular reflections, fluid artifacts, and rapid camera motion often force existing methods to compromise between tracking robustness and reconstruction fidelity. We introduce EndoSplaTAM, a novel dense SLAM framework leveraging 3D Gaussian Splatting (3DGS) to enable high-frequency, photorealistic tissue reconstruction from unposed RGB-D endoscopic streams. Unlike computationally intensive neural implicit representations, our approach utilizes a scalable sub-map strategy that keeps GPU memory and per-frame optimization bounded while maintaining local geometric consistency during long-term exploration. We further propose an artifact-resilient frame-to-model tracking objective combining dense photometric and geometric constraints. Experiments on the C3VD dataset demonstrate that EndoSplaTAM establishes a new state-of-the-art, surpassing representative dense baselines by >10 dB in PSNR while achieving sub-millimeter tracking accuracy and maintaining real-time rendering at >100 FPS. Our method offers a practical leap forward for surgical navigation and augmented reality, showcasing strong potential for high-precision navigation.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/EndoLINA_001.pdf
SharedIt Link: Not yet available
SpringerLink (DOI): Not yet available
Supplementary Material: Not Submitted
Link to Open Review
Open Review Page: https://openreview.net/forum?id=pV6TW6TO1N
BibTex
@InProceedings{MenZih_EndoSplaTAM_MICCAISAT2026,
author = { Men, Zihan AND Wu, Yanhong AND Cheung, Jason Pui Yin AND Wang, Siyu AND Wei, Yuxin AND Zhao, Moxin AND Li, Dawei AND Zhang, Teng AND Meng, Nan},
title = { { EndoSplaTAM: Photorealistic Dense SLAM for Endoscopy Using Gaussian Splatting } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026 Workshops and Challenges},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS 17275},
month = {pending},
page = {pending}
}