13th Conference on Industrial Computed Tomography, Wels, Austria (iCT 2024), www.ict-conference.com/2024 Copyright 2024 - by the Authors. Licensed under a Creative Commons Attribution 4.0 International License. Spatial Resolution-Optimized Artifact Reduction by Combining Dual Energy X-Ray Computed Tomography Data using a Frequency Split Philip Trapp 1,2 , Frederic Ballach 2 , Raoul Christoph 2 , Ralf Christoph 2 , and Marc Kachelrieß 1 1 German Cancer Research Center (DKFZ), Im Neuenheimer Feld 280, 69120 Heidelberg, Germany 2 Werth Messtechnik GmbH, Siemensstraße 19, 35394 Giessen, Germany Corresponding author: philip.trapp@dkfz.de Abstract The measurability of multi-material workpieces with metals using x-ray computed tomography (CT) is often severely limited due to image artifacts. In this paper a new method for artifact reduction in such CT measurements is presented. This method is based on the well-known dual-energy CT (DECT) approach and works solely in image domain. In contrast to conventional DECT, however, our method allows datasets with different spatial resolution values to be combined, leading to a result volume with optimized spatial resolution, as well as noise and artifact characteristics. Compared to the state of the art, the algorithm presented here thus results in a higher precision of the measurement results and/or to a reduction of the measurement time. Keywords: artifact reduction, dual-energy CT, measurement time reduction, multi-material 1 Introduction In CT measurements, beam hardening artifacts occur due to the stronger absorption of low-energy photons compared to photons with higher energy. These artifacts are particularly common in measurements of workpieces with dense materials (e.g. metals) and can lead to strong systematic measurement deviations that severely limit the applicability of x-ray CT for inspection and dimensional measurements of such workpieces. To reduce beam hardening artifacts, strong prefilters are often used to narrow the x-ray spectrum towards being less polychromatic (i.e. closer to monochromatic). One disadvantage of strong prefiltering is that one must accept a reduced contrast-to-noise ratio (CNR), given that the x-ray tube power and the exposure time are constant. Since even strong prefiltering is often not able to reduce beam hardening artifacts to an acceptable level, the DECT technique was suggested many years ago to further reduce these artifacts [1]. In DECT, the object is measured twice (either simultaneously or subsequently) with different photon spectra in order to be able to combine the data thus acquired in an artifact-reducing manner. A very simple and widely used method of DECT artifact reduction performs a linear combination in image domain, which, however, often leads to a strong CNR reduction in the resulting volume [2, 3]. To compensate for the disadvantage of reduced CNRs in DECT artifact reduction without increasing the measurement time there are several options, such as increasing the tube power, detector binning, or software smoothing of the acquired data. However, each of these options may have drawbacks: for example, higher x-ray tube power usually results in larger focal spots, which can have a negative impact on the expected spatial resolution of the CT measurement. Similarly, higher detector binning or smoothing of the measured data leads to a reduced spatial resolution, as well. The frequency split DECT (FSDECT) method proposed here is an improvement of the established DECT method for artifact reduction. In this method, a frequency split procedure [4, 5] is used to extend DECT to datasets with different spatial resolution characteristics so that the resulting dataset has both, reduced artifact content and high spatial resolution. FSDECT can thus be used to counteract the problem of low CNRs caused by DECT without compromising on spatial resolution. Since a distinction is usually made in metrology between metrological resolution and structural resolution, and this work always refers to the structural resolution, the term structural resolution will from now on be used instead of spatial resolution. 2 Materials and Methods 2.1 Working Principle of FSDECT The FSDECT algorithm consists of three steps: First, the modulation transfer function (MTF) of the datasets measured with different x-ray spectra and structural resolutions is matched so that both volumes fHi and fLo have the same structural resolution. The MTF matching is achieved by low-pass filtering the high-resolution volume, which is typically the volume emerging from the low-energy photon spectrum scan (fLo). The second step is a conventional linear combination in image domain [2]. The combined volume fα is thus given by  α (, , ) =  ∙ (LP  Lo (, , )) + (1 − ) ∙  Hi (, , ). (1) In this equation, LP denotes the low-pass filtering operation that maps the MTF of fLo to that of fHi. The linear factor α is set in such a way that the artifacts in fα are reduced to a minimum. As shown in reference [2], this can be done automatically by More info about this article: https://www.ndt.net/?id=29274 e-Journal of Nondestructive Testing - ISSN 1435-4934 - www.ndt.net https://creativecommons.org/licenses/by/4.0/ https://doi.org/10.58286/29274