Supplementary Material: Use of fast realistic simulations on GPU to extract CAD models from microtomographic data in the presence of strong CT artefacts

Supporting material for our paper published by Elsevier in Precision Engineering, see https://doi.org/10.1016/j.precisioneng.2021.10.014 for the paper

To run our code, you must install:

1. Python
2. A C++ compiler with CMake
3. gVirtualXRay (code provided)

Python dependencies:
1. numpy
2. pandas
3. imageio
4. scikit-image
5. sklearn
6. scipy
7. SimpleITK
8. OpenCV
9. cma
10. tifffile import
11. tigre/tomopy (optional)

Contents

1. code:
- gVirtualXRay-1.1.5-Source.zip: Source code of gVirtualXRay
- tutorial.py: the Python code of the registration pipeline in 2D
- lsf.py: Detector response
- utils.py: needed by pseudo3d.py
- pseudo3d.py: Pseudo 3D registration
2. data: contains the input sinograms
- proj-33keV.tif: projections after flatfield correction. Projections are of 1024 pixels. Pixel spacing: 1.9um. 900 angles over 180 degrees. Primary energy : 33 keV. Source-object distance: 145m. Object-detector distance: 80mm.
- proj-0000-80keV.tif: projections of the first slice after flatfield correction. Projections are of 2702 pixels. Pixel spacing: 1.22um. 1750 angles over 360 degrees. Primary energy: 80 keV. Source-object distance: 92m. Object-detector distance: 270mm.
- proj-1023-80keV.tif: projections of the last slice after flatfield correction.
3. Jupyter_Notebooks: contains our tutorial and some videos
- From_CT_acquisition_to_CAD_models.ipynb: Jupyter Notebook in Python
- From_CT_acquisition_to_CAD_models.pdf: Jupyter Notebook as a PDF file
- cube_registration.mp4: successive best results during the optimisation of the matrix properties
- fibre1_registration.mp4: successive best results during the optimisation of the fibre and core radii (before recentring)
- fibre3_registration.mp4: successive best results during the optimisation of the fibre and core radii (after recentring)
- spectrum_registration.mp4: successive best results during the optimisation of the beam spectrum
- laplacian1_registration.mp4: successive best results during the optimisation of the phase contrast and of the fibre and core radii
- laplacian2_registration.mp4: successive best results during the optimisation of the phase contrast and the response of the detector
4. results: registered images
- output33keV: 2D registration
- output80keV: pseudo-3D registration
5. user_study:
- online-form.pdf: the online form
- responses.csv: answers from volunteers
6. visualisations: contains our interactive parallel coordinate plots and a video showing how to use them
- how_to_use_visualisations.mp4: video showing how we used it to analyse the data
- parallel_coordinates_all_data.html: interactive parallel coordinate plots of the fibre and core raii, and the linear attenuation coefficients of the cores, fibres and matrix
- parallel_coordinates-ZNCC.htmll: interactive parallel coordinate plots of the successive ZNCC values during a registration

本数据集为发表于《精密工程》杂志的论文的辅助材料，详见https://doi.org/10.1016/j.precisioneng.2021.10.014获取论文。为运行我们的代码，您必须安装以下软件和库：

1. Python
2. 配备CMake的C++编译器
3. gVirtualXRay（代码已提供）

Python依赖项：
1. numpy
2. pandas
3. imageio
4. scikit-image
5. sklearn
6. scipy
7. SimpleITK
8. OpenCV
9. cma
10. tifffile导入
11. tigre/tomopy（可选）

数据集内容如下：

1. 代码：
- gVirtualXRay-1.1.5-Source.zip：gVirtualXRay的源代码
- tutorial.py：2D配准流程的Python代码
- lsf.py：探测器响应
- utils.py：pseudo3d.py所需
- pseudo3d.py：伪3D配准
2. 数据：包含输入的sinogram
- proj-33keV.tif：经平坦场校正后的投影。投影分辨率为1024像素。像素间距：1.9微米。900个角度覆盖180度。主能量：33千电子伏。源-物体距离：145米。物体-探测器距离：80毫米。
- proj-0000-80keV.tif：经平坦场校正后的第一层投影。投影分辨率为2702像素。像素间距：1.22微米。1750个角度覆盖360度。主能量：80千电子伏。源-物体距离：92米。物体-探测器距离：270毫米。
- proj-1023-80keV.tif：经平坦场校正后的最后层投影。
3. Jupyter_Notebooks：包含我们的教程和一些视频
- From_CT_acquisition_to_CAD_models.ipynb：Python的Jupyter Notebook
- From_CT_acquisition_to_CAD_models.pdf：Jupyter Notebook的PDF文件
- cube_registration.mp4：优化矩阵属性过程中的连续最佳结果
- fibre1_registration.mp4：优化纤维和芯半径（在重新中心之前）过程中的连续最佳结果
- fibre3_registration.mp4：优化纤维和芯半径（在重新中心之后）过程中的连续最佳结果
- spectrum_registration.mp4：优化束谱过程中的连续最佳结果
- laplacian1_registration.mp4：优化相位对比度以及纤维和芯半径过程中的连续最佳结果
- laplacian2_registration.mp4：优化相位对比度和探测器响应过程中的连续最佳结果
4. 结果：配准图像
- output33keV：2D配准
- output80keV：伪3D配准
5. 用户研究：
- online-form.pdf：在线表格
- responses.csv：志愿者的回答
6. 可视化：包含我们的交互式并行坐标图和展示如何使用它们的视频
- how_to_use_visualisations.mp4：展示如何使用它来分析数据的视频
- parallel_coordinates_all_data.html：纤维和芯半径，以及核心、纤维和基质的线性衰减系数的交互式并行坐标图
- parallel_coordinates-ZNCC.htmll：注册过程中连续ZNCC值的交互式并行坐标图