mlsanityData to 14 Oct 2026

UBone3D: Physics-Rectified Conditional Flow Matching for Anatomical 3D Shape Completion from Ultrasound

arXiv 2609.11506 ↗First posted 10 Sep 2026cs.CV eess.IVcited by 0

Not on the map. Why?

Abstract (arXiv)

Three-dimensional ultrasound (US) is a safe, radiation-free complementary modality to CT and X-rays for longitudinal monitoring, yet its segmentation-derived partial point clouds are extremely artifact-laden. Consequently, it is challenging to recover a clean and complete anatomical structure from such US point clouds. In this paper, we present UBone3D, a novel framework based on physics-rectified conditional flow matching (CFM) that performs point cloud completion directly from partial US observations. UBone3D models deterministic physics artifacts (e.g., surface thickening, streaking, dropouts) via a simulated physics proxy, and introduces test-time physics rectification to steer the shape completion. At inference, the completion is jointly steered by two decoupled forces: (1) anatomical plausibility enforced by a CT-trained generative shape prior, BoneFM, and (2) physics consistency enforced by USimNet in the ultrasound formation space. Extensive experiments on simulated and in-vivo data demonstrate significant improvements in reconstruction accuracy and anatomical fidelity over existing baselines.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 33 of its 49 references that are in the map. First posted on arXiv 10 Sep 2026.

Builds on

33 of 49 references.

References · 33

  1. 2020Denoising Diffusion Implicit Modelscited by 9,781
  2. 2022Flow Matching for Generative Modelingcited by 5,702
  3. 2017PointNet++: Deep Hierarchical Feature Learning on Point Sets in a Metric Spacecited by 4,401
  4. 2022Flow Straight and Fast: Learning to Generate and Transfer Data with Rectified Flowcited by 3,700
  5. 2015ShapeNet: An Information-Rich 3D Model Repositorycited by 2,897
  6. 2017FiLM: Visual Reasoning with a General Conditioning Layercited by 2,333
  7. 2022Building Normalizing Flows with Stochastic Interpolantscited by 995
  8. 2022LION: Latent Point Diffusion Models for 3D Shape Generationcited by 553
  9. 2018PCN: Point Completion Networkcited by 384
  10. 2022PhysDiff: Physics-Guided Human Motion Diffusion Modelcited by 368
  11. 2021PoinTr: Diverse Point Cloud Completion with Geometry-Aware Transformerscited by 286
  12. 2020GRNet: Gridding Residual Network for Dense Point Cloud Completioncited by 174
21 more
  1. 2021SnowflakeNet: Point Cloud Completion by Snowflake Point Deconvolution with Skip-Transformercited by 164
  2. 2024Physics-Informed Diffusion Modelscited by 90
  3. 2025PhysTwin: Physics-Informed Reconstruction and Simulation of Deformable Objects from Videoscited by 87
  4. 2022SeedFormer: Patch Seeds based Point Cloud Completion with Upsample Transformercited by 83
  5. 2024Optimal Flow Matching: Learning Straight Trajectories in Just One Stepcited by 78
  6. 2023Aligning Optimization Trajectories with Diffusion Models for Constrained Design Generationcited by 52
  7. 2021CTSpine1K: A Large-Scale Dataset for Spinal Vertebrae Segmentation in Computed Tomographycited by 46
  8. 2023ProxyFormer: Proxy Alignment Assisted Point Cloud Completion with Missing Part Sensitive Transformercited by 43
  9. 2025PhysAnimator: Physics-Guided Generative Cartoon Animationcited by 42
  10. 2023SVDFormer: Complementing Point Cloud via Self-view Augmentation and Self-structure Dual-generatorcited by 34
  11. 2023Ultra-NeRF: Neural Radiance Fields for Ultrasound Imagingcited by 27
  12. 2025SuperPC: A Single Diffusion Model for Point Cloud Completion, Upsampling, Denoising, and Colorizationcited by 21
  13. 2024PCDreamer: Point Cloud Completion Through Multi-view Diffusion Priorscited by 16
  14. 2024Neural Implicit Surface Reconstruction of Freehand 3D Ultrasound Volume with Geometric Constraintscited by 15
  15. 2024DMF-Net: Image-Guided Point Cloud Completion with Dual-Channel Modality Fusion and Shape-Aware Upsampling Transformercited by 5
  16. 2024RoCoSDF: Row-Column Scanned Neural Signed Distance Fields for Freehand 3D Ultrasound Imaging Shape Reconstructioncited by 5
  17. 2025Rethinking Multimodal Point Cloud Completion: A Completion-by-Correction Perspectivecited by 4
  18. 2024Shape Completion in the Dark: Completing Vertebrae Morphology from 3D Ultrasoundcited by 3
  19. 2024Digging into Intrinsic Contextual Information for High-fidelity 3D Point Cloud Completioncited by 3
  20. 2025UltrON: Ultrasound Occupancy Networkscited by 3
  21. 2025US-X Complete: A Multi-Modal Approach to Anatomical 3D Shape Recoverycited by 3

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2025TUS-REC2024: A Challenge to Reconstruct 3D Freehand Ultrasound Without External Tracker
  2. 2026ZODIAC: Zero-shot Octree-based Diffusion for Anatomical Completion
  3. 2026MedPCFM: Improving Medical Point Cloud Completion by Integrating Point Transformers and Flow Matching
  4. 2025Surf2CT: Cascaded 3D Flow Matching Models for Torso 3D CT Synthesis from Skin Surface
  5. 2024NeRF-US: Removing Ultrasound Imaging Artifacts from Neural Radiance Fields in the Wild
  6. 2025UltrasODM: A Dual Stream Optical Flow Mamba Network for 3D Freehand Ultrasound Reconstruction
  7. 2026UltraWorld: Learning Interactive Ultrasound World Models from Untracked Clinical Videos with Acoustic Sampling Map
  8. 2024Fine-grained Context and Multi-modal Alignment for Freehand 3D Ultrasound Reconstruction