xvr
收藏资源简介:
xvr数据集是由麻省理工学院等多个机构创建的,用于训练特定于患者的神经网络,以实现术中X射线与术前3D体积的快速准确注册。该数据集包含了来自三个医院66位患者的多模态影像数据和多种解剖结构的注册数据。数据集通过物理基础的仿真,从患者的术前3D影像中生成高质量的合成X射线训练数据,以克服监督模型在新患者和程序上泛化能力不足的问题。
The XVR dataset was developed by multiple institutions including the Massachusetts Institute of Technology (MIT) for training patient-specific neural networks to achieve fast and accurate registration between intraoperative X-rays and preoperative 3D volumes. This dataset contains multimodal imaging data and registration data for various anatomical structures from 66 patients across three hospitals. It generates high-quality synthetic X-ray training data from patients' preoperative 3D images via physics-based simulation, aiming to overcome the insufficient generalization ability of supervised models on unseen patients and clinical procedures.

- 1Rapid patient-specific neural networks for intraoperative X-ray to volume registration麻省理工学院 · 2025年



