ReferenceQuery Predicted pose Pose distribution Reference Query Predicted pose Pose distribution Figure 1. Given as input a single reference view of a novel object, our method predicts the relative 3D pose of a query view and its ambiguities. We visualize the predicted pose by rendering the object from this pose, but the 3D model is only used for visualization purposes, not as input to our method. Our method works by estimating a probability distribution over the space of 3D poses, visualized here on a sphere centered on the object. We use the canonical pose of the 3D model to visualize this distribution, but not as input to our method. From this distribution, we can also identify the pose ambiguities: For example, in the case of the bottle, any pose with the same pitch and roll is possible; in the case of the mug, a range of poses are possible as the handle is not visible in the query image. Our method is also robust to partial occlusions, as shown on the clock hidden in part by a rectangle in the query image.

Compared to single source systems, stereo X-ray CT systems allow acquiring projection data within a reduced amount of time, for an extended field-of-view, or for dual X-ray energies. To exploit the benefit of a dual X-ray system, its acquisition geometry needs to be calibrated. Unfortunately, in modular stereo X-ray CT setups , geometry misalignment occurs each time the setup is changed, which calls for an efficient calibration procedure. Although many studies have been dealing with geometry calibration of an X-ray CT system, little research targets the calibration of a dual cone-beam X-ray CT system. In this work, we present a phantom-based calibration procedure to accurately estimate the geometry of a stereo cone-beam X-ray CT system. With simulated as well as real experiments, it is shown that the calibration procedure can be used to accurately estimate the geometry of a modular stereo X-ray CT system thereby reducing the misalignment artifacts in the reconstruction volumes.

No abstract

Accurate knowledge of the acquisition geometry of a CT scanning system is key for high quality tomographic imaging. Unfortunately, in modular X-ray CT setups, geometry misalignment occurs each time the setup is changed, which calls for an efficient calibration procedure to correct for geometric inaccuracies. Although many studies have been dealing with the calibration of X-ray CT systems, these are often specifically designed for one setup and/or expensive.In this work, we explore the possibilities of a low-cost, easy-tobuild, and modular phantom, constructed from LEGO bricks, which serves as a structure to hold small metal beads, for geometric calibration of a tomographic X-ray system. By estimating the bead coordinates using deep learning, and minimizing center-to-center distances of the metal beads between measured and reference projection data, geometry parameters are derived. With simulated as well as real experiments, it is shown that the LEGO phantom can be used to accurately estimate the geometry of a modular X-ray CT system.

An issue in computerized X-ray tomography is the limited size of available detectors relative to objects of interest. A solution was provided in the past two decades by positioning the detector in a lateral offset position, increasing the effective field of view (FOV) and thus the diameter of the reconstructed volume. However, this introduced artifacts in the obtained reconstructions, caused by projection truncation and data redundancy. These issues can be addressed by incorporating an additional data weighting step in the reconstruction algorithms, known as redundancy weighting. In this work, we present an implementation of redundancy weighting in the widely-used Simultaneous Iterative Reconstruction Technique (SIRT), yielding the W-SIRT method. The new technique is validated using geometric phantoms and a rabbit specimen, by performing both simulation studies as well as physical experiments. The experiments are carried out in a highly flexible stereoscopic X-ray system equipped with X-ray image intensifiers (XRIIs). The simulations showed that higher values of CNR could be obtained using the W-SIRT approach as compared to a weighted implementation of SART. The convergence rate of the W-SIRT was accelerated by including a relaxation parameter in the W-SIRT algorithm, creating the aW-SIRT algorithm. This allowed to obtain the same results as the W-SIRT algorithm, but at half the number of iterations, yielding a much shorter computation time. The aW-SIRT algorithm has proven to perform well for both large as well as small regions of overlap, outperforming the pre-convolutional Feldkamp-David-Kress (FDK) algorithm for small overlap regions (or large detector offsets). The experiments confirmed the results of the simulations. Using the aW-SIRT algorithm, the effective FOV was increased by >75%, only limited by experimental constraints. Although an XRII is used in this work, the method readily applies to flat-panel detectors as well.