人工放射线-地标数据集

本研究使用了一个名为人工放射线-地标数据集，该数据集由骨生物工程研究中心创建，用于训练深度神经网络进行地标检测和图像风格转换。数据集包含9751对DRR-地标对，用于训练和测试深度残差网络（ResNet）。数据集的创建过程涉及从女性受试者的颅颈椎CT数据重建3D颅骨模型，并自动生成数字重建放射线（DRRs）和放射线颅骨地标。该数据集主要应用于3D到2D的颅骨模型注册，以解决在双荧光X射线图像中不完全捕捉颅骨区域的问题。

This study employs the Artificial Radiograph-Landmark Dataset, which was developed by the Bone Bioengineering Research Center for training deep neural networks in landmark detection and image style transfer. The dataset consists of 9751 DRR-landmark pairs for training and testing deep residual networks (ResNet). The dataset creation process entails reconstructing 3D skull models from cervicocranial CT data of female subjects, and automatically generating Digitally Reconstructed Radiographs (DRRs) and radiographic skull landmarks. This dataset is primarily applied to 3D-to-2D skull model registration to address the issue of incomplete skull region capture in biplanar fluoroscopic X-ray images.