Single Material Beam Hardening Correction in CT Using an Analytical Energy Response Model

Material decomposition algorithms for dual energy CT require knowledge on the energy response of the used system. This knowledge can be gained either through calibration measurements or using an explicit analytical model. While the former approach is preferred by vendors, it tends to be intimately linked with the chosen image reconstruction approach and is thus not easily amenable to using other, potentially more advanced, image reconstruction methods. Models for the detector response and the source spectrum have been published in the literature. Here, we investigate how satisfactory these general models are to represent the characteristics of one specific scanner. Specifically, the models are tested by using them to perform a water-based, single material beam hardening correction on real CT data. This test is performed using an anthropomorphic head phantom and the ACR CT accreditation phantom both scanned at two different X-ray tube settings that are typically used for dual energy CT. The scans are performed using a state-of-the-art diagnostic CT scanner (SOMATOM Drive, Siemens Healthcare GmbH, Germany). The images corrected for beam hardening using the models are compared with the classical reconstructions provided by the vendor, for which beam hardening correction is based on real measurements from a calibration phantom. The study demonstrates that the results provided by the vendor can be reproduced with high accuracy (within 1-3 HU in mean value) provided some analytical correction for forward scatter is included as well.