LUDO: Low-Latency Understanding of Deformable Objects Using Point Cloud Occupancy Functions
Pit Henrich, Franziska Mathis-Ullrich, Paul Maria Scheikl
AI summary
Problem
Intraoperative tissue deformation renders preoperative imaging inaccurate, yet real-time 3D structural localization remains critical for safe medical interventions like biopsies.
Approach
The method employs a trained occupancy network conditioned on a single-view point cloud to instantly generate a dense 3D representation of both external and internal anatomy, supplemented by uncertainty quantification and feature explainability.
Key results
- 98.9% success rate in real-world robotic puncturing of internal targets
- Sub-30 ms inference time for dense 3D structural reconstruction
- Superior ROI localization accuracy and lower memory usage compared to baselines
- Integrated uncertainty estimation and input feature explainability for safety
Why it matters
Provides a fast, reliable alternative to deformable registration for real-time surgical planning and robotic intervention in soft tissue.
Abstract
Accurately determining the shape of deformable objects and the location of their internal structures is crucial for medical tasks that require precise targeting, such as robotic biopsies. We introduce LUDO, a method for accurate low-latency understanding of deformable objects. LUDO reconstructs objects in their deformed state, including their internal structures, from a single-view point cloud observation in under 30 ms using occupancy networks. LUDO provides uncertainty estimates for its predictions. Additionally, it provides explainability by high- lighting key features in its input observations. Both uncertainty and explainability are important for safety-critical applications such as surgery. We evaluate LUDO in real-world robotic experiments, achieving a success rate of 98.9% for puncturing various regions of interest (ROIs) inside deformable objects. We compare LUDO to a popular baseline and show its superior ROI localization accuracy, training time, and memory requirements. LUDO demonstrates the potential to interact with deformable objects without the need for deformable registration methods.