PwDECT: Dual Energy CT with local SNR-weights for improved reconstruction of objects with radially anisotropic transmissions

Abstract

For objects which are either made from a single or multiple materials, CT reconstruction assumes more or less implicitly the equivalence of the different X-rays which are passing through the object. Equivalence in terms of energy and hence attenuation (or refraction in the case of phase-contrast scanners) contradicts the polychromaticity of the X-ray spectrum, causing beam hardening artifacts in the resulting volume images. Meanwhile, equivalence in terms of transmission (which favors an optimal signal-to-noise ratio SNR) is often ignored. CT scans which comprise very different transmissions (either because of anisotropic shape or anisotropic materials distribution) feature strongly anisotropic volume image noise or even metal artifacts for strongly varying material densitites. For planar objects transmission can be equalized by replacing CT with Computed Laminography (CL), yet CL is only reconstructing partial volume information while eclipsing structural in-plane components. This work aims at combining the information from two or more CT scans which were recorded with different X-ray spectra (commonly referred to as Dual Energy CT) in order to collect strong signals„ both along short and long path lengths, while avoiding beam hardening during reconstruction. The method is demonstrated for impact damaged glass-fiber boards and for the circuit board in a smartphone.