Data for paper: "Data-driven image mechanics (D2IM): a deep learning approach to predict displacement and strain fields from undeformed X-ray tomography images - Evaluation of bone mechanics"

Data used in the paper: "Data-driven image mechanics (D2IM): a deep learning approach to predict displacement and strain fields from undeformed X-ray tomography images - Evaluation of bone mechanics"by Peter Soar, Marco Palanca, Enrico Dall'Ara and Gianluca Tozzi<br>submitted to publication in Extreme Mechanical Letters, 2024<br>There are 2 parts of this directory:The main dataset 'D2IM_Data':<br>- 'input' comprises slices of XTC scans of porcine vertebra. There is a subfolder 'Removed' containing slices that were manually removed from the dataset due to not containing enough bone. The original 3D scans of the images can be found at: https://orda.shef.ac.uk/articles/dataset/Data_for_paper_MicroFE_models_of_porcine_vertebrae_with_induced_bone_focal_lesions_validation_of_predicted_displacements_with_Digital_Volume_Correlation_/16732441/1<br>- 'mask' is another input that was generated from these scan slices. <br>' 'Target' contains three subfolders of the displacements ('U','V','W'), used as the target/ground truth. These fields were obtained using the DIC/DVC software SPAM.<br>'Clinical' folder contains a few examples used for testing the sensitivity of the D2IM model to lower resolution input. Clinical images were collected at the CT Scan Department, Northern General Hospital, Sheffield.The 'Test Predictions' directory has three archives, each containing images showing the predicted displacements and displacement errors for the full set of test data. Each archive corresponds to the named displacement component u, v or w.A copy of the code for the D2IM prototype model can be found on GitHub at: https://github.com/PeterSoar/D2IM_PrototypeFor any questions the reader is encouraged to contact Dr Peter Soar ([e-mail]).

本数据集对应论文《数据驱动图像力学（D2IM）：从未变形X射线断层扫描图像预测位移与应变场的深度学习方法——骨力学评估》（Data-driven image mechanics (D2IM): a deep learning approach to predict displacement and strain fields from undeformed X-ray tomography images - Evaluation of bone mechanics），作者为Peter Soar、Marco Palanca、Enrico Dall'Ara与Gianluca Tozzi，2024年提交至《极端力学快报（Extreme Mechanical Letters）》待发表。
本数据集目录包含两部分：主数据集「D2IM_Data」：
- 「input」文件夹存储猪椎体的X射线断层扫描（X-ray Tomography, XTC）切片，其中设有子文件夹「Removed」，存放因骨骼占比不足而被人工剔除的扫描切片。原始三维扫描图像可通过以下链接获取：https://orda.shef.ac.uk/articles/dataset/Data_for_paper_MicroFE_models_of_porcine_vertebrae_with_induced_bone_focal_lesions_validation_of_predicted_displacements_with_Digital_Volume_Correlation_/16732441/1
- 「mask」为另一类输入数据，由上述扫描切片生成。
「Target」文件夹包含三个子文件夹，分别对应位移分量U、V、W，作为模型训练的目标真值（ground truth）。上述位移场通过数字图像相关（Digital Image Correlation, DIC）/数字体相关（Digital Volume Correlation, DVC）软件SPAM计算得到。
「Clinical」文件夹包含少量示例样本，用于测试D2IM模型对低分辨率输入的敏感性。临床图像采集自英国谢菲尔德北部总医院CT扫描科。
「Test Predictions」目录包含三个归档文件，每个归档均存储展示全测试集预测位移及位移误差的图像，分别对应位移分量u、v、w。
D2IM原型模型的代码副本可在GitHub获取：https://github.com/PeterSoar/D2IM_Prototype
如有任何疑问，欢迎联系Peter Soar博士（邮箱：[e-mail]）。