Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Evaluation of Intra-operative Patient-specific Methods for Point Cloud Completion for Minimally Invasive Liver Interventions

Nakul Poudel, Zixin Yang, Kelly Merrell, Richard Simon, Cristian A. Linte

TL;DR

The paper tackles the critical problem of unstable 3D-3D registration in image-guided liver surgery caused by partial, noisy intra-operative surfaces. It systematically evaluates six point cloud completion methods (FoldingNet, PCN, TopNet, GRNet, PoinTr, AdaPoinTr) on a large, patient-specific liver dataset under canonical, non-canonical, and noisy conditions. AdaPoinTr consistently provides the best completion quality in canonical poses, while all methods show substantial performance degradation in non-canonical poses and under noise, with PoinTr offering robustness to higher noise levels. The findings motivate developing rotation-equivariant and noise-resilient PCC approaches and integrating PCC more tightly into registration pipelines for reliable image-guided liver interventions.

Abstract

The registration between the pre-operative model and the intra-operative surface is crucial in image-guided liver surgery, as it facilitates the effective use of pre-operative information during the procedure. However, the intra-operative surface, usually represented as a point cloud, often has limited coverage, especially in laparoscopic surgery, and is prone to holes and noise, posing significant challenges for registration methods. Point cloud completion methods have the potential to alleviate these issues. Thus, we explore six state-of-the-art point cloud completion methods to identify the optimal completion method for liver surgery applications. We focus on a patient-specific approach for liver point cloud completion from a partial liver surface under three cases: canonical pose, non-canonical pose, and canonical pose with noise. The transformer-based method, AdaPoinTr, outperforms all other methods to generate a complete point cloud from the given partial liver point cloud under the canonical pose. On the other hand, our findings reveal substantial performance degradation of these methods under non-canonical poses and noisy settings, highlighting the limitations of these methods, which suggests the need for a robust point completion method for its application in image-guided liver surgery.

Evaluation of Intra-operative Patient-specific Methods for Point Cloud Completion for Minimally Invasive Liver Interventions

TL;DR

The paper tackles the critical problem of unstable 3D-3D registration in image-guided liver surgery caused by partial, noisy intra-operative surfaces. It systematically evaluates six point cloud completion methods (FoldingNet, PCN, TopNet, GRNet, PoinTr, AdaPoinTr) on a large, patient-specific liver dataset under canonical, non-canonical, and noisy conditions. AdaPoinTr consistently provides the best completion quality in canonical poses, while all methods show substantial performance degradation in non-canonical poses and under noise, with PoinTr offering robustness to higher noise levels. The findings motivate developing rotation-equivariant and noise-resilient PCC approaches and integrating PCC more tightly into registration pipelines for reliable image-guided liver interventions.

Abstract

The registration between the pre-operative model and the intra-operative surface is crucial in image-guided liver surgery, as it facilitates the effective use of pre-operative information during the procedure. However, the intra-operative surface, usually represented as a point cloud, often has limited coverage, especially in laparoscopic surgery, and is prone to holes and noise, posing significant challenges for registration methods. Point cloud completion methods have the potential to alleviate these issues. Thus, we explore six state-of-the-art point cloud completion methods to identify the optimal completion method for liver surgery applications. We focus on a patient-specific approach for liver point cloud completion from a partial liver surface under three cases: canonical pose, non-canonical pose, and canonical pose with noise. The transformer-based method, AdaPoinTr, outperforms all other methods to generate a complete point cloud from the given partial liver point cloud under the canonical pose. On the other hand, our findings reveal substantial performance degradation of these methods under non-canonical poses and noisy settings, highlighting the limitations of these methods, which suggests the need for a robust point completion method for its application in image-guided liver surgery.

Paper Structure

This paper contains 20 sections, 5 equations, 2 figures, 4 tables.

Table of Contents

  1. INTRODUCTION
  2. Methodology
  3. Problem Definition
  4. FoldingNet
  5. PCN
  6. TopNet
  7. GRNet
  8. PoinTr and AdaPoinTr
  9. Experiments
  10. Intra-operative Liver Point Cloud Data
  11. Point Cloud Normalization
  12. Experimental Setup
  13. Non-canionical pose setup
  14. Noise setup
  15. Evaluation Metrics
  16. ...and 5 more sections

Figures (2)

  • Figure 1: Illustration of eight different partial inputs generated for each liver model during testing. Eight distinct viewpoints are used to create these partial inputs, with missing points located in various regions.
  • Figure 2: Qualitative comparison of point cloud completion results on canonical liver models. The first column shows the input incomplete point clouds, with 25% missing regions in the first row, 50% missing regions in the second, and 75% missing regions in the third. The second column shows the ground truth complete point clouds. The subsequent columns present the completion results from various PCC methods: AdaPoinTr, PoinTr, PCN, GRNet, TopNet, and FoldingNet. The colorbar represents the Euclidean Distance (in mm) from each point in the prediction to its nearest point in the ground truth.