Evaluation of intraoperative patient-specific methods for point cloud completion for minimally invasive liver interventions

2025-03-15

Medical Imaging 2025: Image-Guided Procedures, Robotic Interventions, and Modeling

AI Summary

In laparoscopic liver surgery, incomplete intraoperative point clouds—characterized by extensive holes and high noise—severely hinder accurate registration between preoperative models and intraoperative data. To address this, this work presents the first systematic evaluation of six state-of-the-art point cloud completion methods under clinically relevant conditions: canonical/non-canonical poses and additive noise, all within patient-specific liver scenarios. We propose AdaPoinTr, a Transformer-based model incorporating geometric modeling and patient-specific shape priors, and introduce a novel quantitative evaluation framework integrating both geometric fidelity and anatomical plausibility. Results show AdaPoinTr achieves top performance under canonical poses; however, all methods exhibit significant degradation under non-canonical poses and noise—revealing a critical robustness bottleneck in real surgical settings. This study underscores the urgent need for pose-invariant and noise-resilient point cloud completion for clinical deployment, providing empirical evidence and concrete directions for developing robust intraoperative registration pipelines.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Intelligent Robots: State Estimation
• Natural Language Processing: Safety and Robustness

Application Category

• Graph Algorithms and Modeling for the Web: Algorithms and analysis for incomplete, noisy, or partially observed Web-related graphs
• User Modeling, Personalization and Recommendation: Accountability, Transparency, and Ethics for personalization
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics

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.

Problem

• Evaluates point cloud completion methods for liver surgery.
• Focuses on patient-specific liver surface reconstruction.
• Identifies limitations in non-canonical poses and noisy settings.

Innovation

• Evaluates six point cloud completion methods
• Focuses on patient-specific liver surface completion
• Identifies AdaPoinTr as best for canonical poses