Projection-angle-dependent distortion correction in high-speed image-intensifier-based x-ray computed tomography

Geometric distortion is inevitable in facilities using x-ray image intensifiers. When the induced distortion pattern varies over time, each recorded frame should be corrected accordingly, which is the case in conventional C-arm imaging, for example. This demonstrates the need for reliable and easy-to-use, projection-angle-dependent correction methods. In the present work, we demonstrate such a dynamic approach, based on digital image correlation (DIC). We validate the method in a set-up for high-speed x-ray tomography, where the variable distortion is induced by an inhomogeneous distribution of ferromagnetic components in the sample rotation stage. By comparing the ideal positions of metal beads in a rectilinear pattern and their positions in the corrected radiographs of that pattern, we deduced the minimum number of frames required to estimate the varying distortion behavior during a full revolution of the stage. Next, this method was validated in a geometry calibration algorithm for a tomographic set-up, as well as in a tomographic reconstruction. Before the application of any distortion correction, the recorded images suffer from, on average, a mean and maximum distortion of 11.12 pixels (1.56 mm) and 42.79 pixels (6.10 mm), respectively. From our experiments, we conclude that three projections, sampled with a 120 degrees interval, are sufficient to correct any frame recorded under an intermediate angle, with mean and maximum residual errors respectively below 0.48 pixels (0.068 mm) and 1.68 pixels (0.240 mm) while using a 14 image intensifier covering a 292 mm x 292 mm field of view. These results imply a decrease of the mean and maximum distortion errors of at least 96%, regardless of the projection angle. Next, the results showed improved accuracy in the system's geometry calibration, which resulted in a reduced blurring of edges and better contrast, as well as shape preservation in the tomographic reconstruction. This demonstrates that the new method is accurate and reliable and, due to the common availability of DIC software, it is very accessible and easy to use.

Automated summary

Geometric distortion in x-ray tomography can be effectively corrected by a dynamic method based on digital image correlation. The results demonstrate that three projections with a 120 degrees interval are sufficient to correct any frame, reducing the distortion errors by at least 96%. This new method improves system geometry calibration accuracy and enhances shape preservation in tomographic reconstruction.