Related Experiment Video
Updated: Mar 27, 2026
06:39
Three-dimensional Location Approach with Silk Thread Guided Laparoscopic Segmentectomy for Liver Tumor
Published on: May 23, 2025
546
Evaluation of Intra-operative Patient-specific Methods for Point Cloud Completion for Minimally Invasive Liver
Nakul Poudel1, Zixin Yang1, Kelly Merrell1
1Center for Imaging Science, Rochester Institute of Technology, Rochester, NY 14623, USA.
Summary
This study evaluated point cloud completion methods for image-guided liver surgery. AdaPoinTr excelled in canonical poses but struggled with non-canonical poses and noise, indicating a need for more robust solutions.
Area of Science:
- Medical Imaging
- Computer-Aided Surgery
- Computational Geometry
Background:
- Accurate registration of pre-operative models and intra-operative surfaces is vital for image-guided liver surgery.
- Intra-operative surface data, often a partial, noisy point cloud, presents challenges for registration, especially in laparoscopic settings.
Purpose of the Study:
- To evaluate state-of-the-art point cloud completion methods for improving intra-operative liver surface data.
- To identify the optimal completion method for patient-specific liver point cloud reconstruction.
Main Methods:
- Investigated six advanced point cloud completion techniques.
- Assessed performance on patient-specific liver data under canonical, non-canonical, and noisy pose conditions.
Main Results:
- The transformer-based AdaPoinTr method demonstrated superior performance for complete point cloud generation in canonical poses.
- Significant performance degradation was observed for all methods in non-canonical poses and noisy environments.
Conclusions:
- While AdaPoinTr shows promise, current point cloud completion methods have limitations for real-world image-guided liver surgery due to pose variations and noise.
- Development of robust point completion algorithms is necessary for reliable intra-operative surface reconstruction in liver surgery.
