UBone3D：用于超声解剖三维形状补全的物理校正条件流匹配
UBone3D: Physics-Rectified Conditional Flow Matching for Anatomical 3D Shape Completion from Ultrasound
查看机构详情
- University of Alberta（阿尔伯塔大学）
- ShanghaiTech University（上海科技大学）
机构由 AI 辅助整理，请以论文原文为准。
浏览论文内容
中文总结 AI 辅助
UBone3D利用物理校正条件流匹配，结合CT先验和超声模拟，从含伪影的超声点云中完成解剖形状补全，显著提升重建精度。
中文摘要 AI 辅助
三维超声（US）是一种安全、无辐射的补充模态，可与CT和X射线一起用于纵向监测，然而其分割得到的部分点云含有大量伪影。因此，从这类超声点云中恢复干净且完整的解剖结构具有挑战性。本文提出UBone3D，一种基于物理校正条件流匹配（CFM）的新框架，可直接从部分超声观测中执行点云补全。UBone3D通过模拟物理代理来建模确定性物理伪影（如表面增厚、条纹、缺失），并引入测试时物理校正以引导形状补全。在推理时，补全由两个解耦的力共同引导：（1）由CT训练的生成形状先验BoneFM强加的解剖合理性，以及（2）由USimNet在超声形成空间中强加的物理一致性。在模拟和体内数据上的大量实验表明，与现有基线相比，重建精度和解剖保真度有显著提升。
英文摘要
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.
