Arcadu F, Marone F, Stampanoni M
Institute for Biomedical Engineering, ETH Zurich, 8092 Zurich, Switzerland.
Swiss Light Source, Paul Scherrer Institute, 5232 Villigen, Switzerland.
J Synchrotron Radiat. 2017 Jan 1;24(Pt 1):205-219. doi: 10.1107/S1600577516015794.
This paper introduces two novel strategies for iterative reconstruction of full interior tomography (FINT) data, i.e. when the field of view is entirely inside the object support and knowledge of the object support itself or the attenuation coefficients inside specific regions of interest are not available. The first approach is based on data edge-padding. The second technique creates an intermediate virtual sinogram, which is, then, reconstructed by a standard iterative algorithm. Both strategies are validated in the framework of the alternate direction method of multipliers plug-and-play with gridding projectors that provide a speed-up of three orders of magnitude with respect to standard operators implemented in real space. The proposed methods are benchmarked on synchrotron-based X-ray tomographic microscopy datasets of mouse lung alveoli. Compared with analytical techniques, the proposed methods substantially improve the reconstruction quality for FINT underconstrained datasets, facilitating subsequent post-processing steps.
本文介绍了两种用于全内层析成像（FINT）数据迭代重建的新策略，即当视野完全位于物体支撑区域内且无法获得物体支撑区域本身或特定感兴趣区域内的衰减系数时的情况。第一种方法基于数据边缘填充。第二种技术创建一个中间虚拟正弦图，然后通过标准迭代算法进行重建。这两种策略都在乘数交替方向法即带网格投影仪的即插即用框架中得到验证，相对于在实空间中实现的标准算子，该框架可将速度提高三个数量级。所提出的方法在基于同步加速器的小鼠肺泡X射线断层显微镜数据集上进行了基准测试。与分析技术相比，所提出的方法显著提高了FINT欠约束数据集的重建质量，便于后续的后处理步骤。
