Signal-to-thickness calibration and pixel-wise interpolation for beam-hardening artefact reduction in microCT

Abstract

X-ray computed tomography (CT) reconstruction suffers from beam-hardening artefacts caused by the polychromaticity of virtually all lab-based X-ray sources. A method to correct for beam-hardening is a direct, pixel-wise signal-to-thickness calibration (STC). We compare reconstructions of conventionally flat-field corrected as well as STC preprocessed measurements of various samples performed on a commercial microCT device based on a flat-panel detector. We show that a good estimate between the transmission signal and the respective material thickness can be given by multiple exponential functions. We further compare the exponential interpolation approach to a hyperbolic model, which reduces the number of necessary calibration measurements significantly. Our method shows that typical beam-hardening artefacts like cupping and filling can be almost completely suppressed and a significant contrast increase is gained. The method can be applied with little additional calibration and computation effort and allows shorter acquisition times since beam filtration can be reduced or omitted.

A Novel and Reliable Pixel Response Correction Method (DAC-Shifting) for Spectral Photon Counting CT Imaging

“…Even though several papers are available on STC-P, none of them describe the relation between system stability and effectiveness ( [9][10][11]15]). This makes a direct comparison to other papers challenging.…”

“…The median number of photon-counts over the five projections for each thickness was divided by the median counts measured at thickness 0. We subsequently modelled the individual pixel response using the global hyperbolic interpolation method from [15]. For STC-D, one STC calibration curve was fitted for every threshold, using the medians of the detector.…”

Complete elimination of the beam hardening effect in quantitative absorption tomography using polychromatic x‐rays with single‐component specimens

“…A common post-processing approach of BH effect correction involves linearization procedures, in which the deviation from linearity of the relationship between the logarithm of the ratio of intensities and the column thickness is approximated as a fitting function. [27][28][29][30][31] The fitting procedures require experimental measurements of the intensities attenuated by references of known geometry, density (or concentration), and made of the same (or similar) material as that of the specimen. For this technique to work, the calibrating references must be available and the reference data must be taken in identical settings as those of the measurements of the specimen.…”