A Review of 3D Reconstruction of Deformable Tissues in Robotic Surgery

Paper for MICCAI 2024 Workshop: Embodied AI and Robotics for HealTHcare (EARTH)

Related Works

Please follow the readme files of the following 4 models for environment installation and data processing to reproduce their results.

• EndoNeRF: EndoNeRF README
• EndoSurf: EndoSurf README
• LerPlane: LerPlane README
• 4D-GS: 4DGaussians README

Dataset Preperation

1. EndoNeRF dataset: Please fill in this form [Sample Dataset for EndoNeRF] to download EndoNeRF dataset.
2. StereoMIS dataset: Please download the StereoMIS dataset from here, then process the data by Robust Camera Pose Estimation for Endoscopic Videos, and organize it in the structure as EndoNeRF:

   + data1
       |
       |+ depth/           # depth maps
       |+ masks/           # binary tool masks
       |+ images/          # rgb images
       |+ pose_bounds.npy  # camera poses & intrinsics in LLFF format
   

   Stereo depth maps are obtained by STTR-Light. You can also generate monocular depth maps by Dpeth-Anything.
3. C3VD dataset: Please download the C3VD dataset from here and the depth map is included in the dataset.

Acknowledgements

We would like to acknowledge the following excellent works: EndoNeRF(Wang et al.), EndoSurf(Zha et al.), LerPlane(Yang et al.), 4D-GS(Wu et al.).