dTwin4SkullShapes: a Digital Twin Dataset of Human Skull Shapes for Skull Missing Vs. Existing Part Learning and Prediction

This dataset include the total of 757 human skulls reconstructed from computed tomography (CT) image sets. The CT image sets were constructed from the three main databases: (1) The Cancer Imaging Archive (DOI: 10.7937/TCIA.HMQ8-J677), (2) The New Mexico Decedent CT Images (DOI: 10.1055/s-0041-1730999), and (3) The MGU500+ (DOI: 10.1016/j.dib.2021.107524). The selected skull image sets have normal and full skull structures. We employed the voxelization technique to reconstruct skull meshes from CT image slices of each subject. The skull meshes were stored in *.off files. For each skull mesh, we estimated the skull shape by estimating the alpha shape of the skull meshes and stored in *.off files. The feature points on the skulls were also manually picked and stored in *.csv files with the form of 16x3 matrices, in which 16 is the number of feature points. A template skull shape was also deformed to the skull shapes of all subjects and stored in *.off files. The deformed skull shapes have unified skull feature points throughout all subjects. Moreover, we also added the source codes of the project. The details of the dataset were presented in the ReadMe.txt file. More details of how to use the source code project and the dataset were presented in the dTwin4SkullShapeTutorials.pdf file.本数据集共计包含757例由计算机断层扫描（CT）图像集重建的人类颅骨。所用CT图像集来源于三大主流数据库：(1) 癌症影像档案库（DOI: 10.7937/TCIA.HMQ8-J677）；(2) 新墨西哥死者CT图像库（DOI: 10.1055/s-0041-1730999）；(3) MGU500+（DOI: 10.1016/j.dib.2021.107524）。所遴选的颅骨图像集均具备完整且正常的颅骨结构。我们采用体素化技术，从每位受试者的CT图像切片中重建颅骨网格，所有颅骨网格均以*.off格式文件存储。针对每个颅骨网格，我们通过估计其alpha形状来推导颅骨形态，并将该形态结果同样存储为*.off格式文件。我们还手动标注了颅骨上的特征点，以16×3矩阵的形式将其存储于*.csv格式文件中，其中16代表特征点的总数量。此外，我们将一个模板颅骨形态配准至所有受试者的颅骨形态，生成的变形后颅骨形态均以*.off格式文件存储，且所有受试者的变形颅骨均具备统一的颅骨特征点标注。本数据集还附带了项目的源代码，数据集的详细说明已收录于ReadMe.txt文件中；关于源代码项目与数据集的使用方法的更多细节，则收录于dTwin4SkullShapeTutorials.pdf文件中。