EndoGSim: Physics-Aware 4D Dynamic Endoscopic Scene Simulations via MLLM-Guided Gaussian Splatting

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Liu, Changjing, Huang, Yiming, Bai, Long, Cui, Beilei, Ren, Hongliang
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910223819603968
author Liu, Changjing
Huang, Yiming
Bai, Long
Cui, Beilei
Ren, Hongliang
author_facet Liu, Changjing
Huang, Yiming
Bai, Long
Cui, Beilei
Ren, Hongliang
contents In robot-assisted minimally invasive surgery, high-fidelity dynamic endoscopic scene reconstruction and simulation are crucial to enhancing downstream tasks and advancing surgical outcomes. However, existing methods primarily focus on visual reconstruction, lacking physics-based descriptions of the scene required for realistic simulation. We propose a unified framework that achieves physics-aware reconstruction and physical simulation of endoscopic scenes through Multi-modal Large Language Models (MLLMs)-guided Gaussian Splatting. Our approach utilizes 4D Gaussian Splatting (4DGS) integrated with pre-trained segmentation and depth estimation to represent deformable tissues and tools. To achieve automatic inference of physical properties, we introduce an object-wise material field that initializes material parameters via MLLM and refines them through a differentiable Material Point Method (MPM) under joint supervision from rendered images and optical flow. Validated on both open-source and in-house datasets, our framework achieves superior simulation fidelity and physical accuracy compared to state-of-the-art methods, underscoring its potential to advance robot-assisted surgical applications.
format Preprint
id arxiv_https___arxiv_org_abs_2605_16022
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle EndoGSim: Physics-Aware 4D Dynamic Endoscopic Scene Simulations via MLLM-Guided Gaussian Splatting
Liu, Changjing
Huang, Yiming
Bai, Long
Cui, Beilei
Ren, Hongliang
Computer Vision and Pattern Recognition
In robot-assisted minimally invasive surgery, high-fidelity dynamic endoscopic scene reconstruction and simulation are crucial to enhancing downstream tasks and advancing surgical outcomes. However, existing methods primarily focus on visual reconstruction, lacking physics-based descriptions of the scene required for realistic simulation. We propose a unified framework that achieves physics-aware reconstruction and physical simulation of endoscopic scenes through Multi-modal Large Language Models (MLLMs)-guided Gaussian Splatting. Our approach utilizes 4D Gaussian Splatting (4DGS) integrated with pre-trained segmentation and depth estimation to represent deformable tissues and tools. To achieve automatic inference of physical properties, we introduce an object-wise material field that initializes material parameters via MLLM and refines them through a differentiable Material Point Method (MPM) under joint supervision from rendered images and optical flow. Validated on both open-source and in-house datasets, our framework achieves superior simulation fidelity and physical accuracy compared to state-of-the-art methods, underscoring its potential to advance robot-assisted surgical applications.
title EndoGSim: Physics-Aware 4D Dynamic Endoscopic Scene Simulations via MLLM-Guided Gaussian Splatting
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2605.16022