Related Experiment Video
Updated: May 23, 2025
09:10
Digital Hybrid Model Preparation for Virtual Planning of Reconstructive Dentoalveolar Surgical Procedures
Published on: August 5, 2021
1.7K
UW-DNeRF: Deformable Soft Tissue Reconstruction With Uncertainty-Guided Depth Supervision and Local Information
IEEE Transactions on Medical Imaging
|March 11, 2025
Summary
UW-DNeRF enhances deformable soft tissue reconstruction from endoscopic videos using neural radiance fields. It improves accuracy by addressing inaccurate depth priors and incorporating local tissue details for better deformation capture.
Area of Science:
- Medical Imaging
- Computer Vision
- Biomedical Engineering
Background:
- Reconstructing deformable soft tissues from endoscopic videos is challenging.
- Existing methods struggle with inaccurate depth priors and lack of local detail capture.
Purpose of the Study:
- To introduce UW-DNeRF, a novel approach for high-quality deformable tissue reconstruction.
- To improve the accuracy and detail of reconstructions by addressing limitations in current methods.
Main Methods:
- Utilizing neural radiance fields for reconstruction.
- Implementing an uncertainty-guided depth supervision strategy to handle inaccurate depth information.
- Employing a local window-based information sharing scheme for enhanced detail and deformation capture.
Main Results:
- UW-DNeRF demonstrates superior performance compared to state-of-the-art methods.
- The method achieves high-quality reconstruction of deformable tissues.
- Improved capture of local details and tissue deformations.
Conclusions:
- UW-DNeRF offers a significant advancement in deformable soft tissue reconstruction from endoscopic videos.
- The proposed uncertainty-guided depth supervision and local information sharing effectively address key challenges in the field.
