中文NeuralBoneReg：基于点云的实例特定无标签多模态骨表面配准方法

在计算机辅助骨科手术中，术前影像制定的个性化计划需精准配准到术中数据。针对不同模态间骨表面配准困难的问题，本文提出一种鲁棒、自动、与模态无关的配准方法，无需手动干预，可提升手术精度与安全性。

arXiv:2511.14286v2 Announce Type: replace Abstract: In computer- and robot-assisted orthopedic surgery (CAOS), patient-specific surgical plans derived from preoperative imaging define target locations and implant trajectories. During surgery, these plans must be accurately transferred, relying on precise cross-registration between preoperative and intraoperative data. However, substantial modality heterogeneity across imaging modalities makes this registration challenging and error-prone. Robust, automatic, and modality-agnostic bone surface registration is therefore clinically important. We p