Potentials and limitations of simulation based artefact correction in computed tomography

In computed Tomography common artefact correction methods generally address one type of artefacts, thus in praxis usually a combination of different methods is used. Without expert knowledge the user of 3D X-ray scanners is frequently over-challenged figuring out what kind of correction method should be used and when. Simulation based artefact correction in contrast is capable of dealing with multiple kinds of artefacts simultaneously. This includes e.g. beam hardening, partial volume artefacts, off-focal radiation and scatter. The simulation based approach uses prior knowledge about the specimen and the X-ray parameters for calculation of artificial projection images. While in medicine the structure of the specimen (patient) is generally unknown, in field of none destructive testing and quality assurance specimens often exists in form of design data. During simulation based artefact correction artificial images are calculated for every measured projection. Aim is the identification and correction of corrupted projection integrals. A common example is scatter correction, where scatter distribution is determined by Monte-Carlo simulation. For correction the scatter is subtracted from measured projection data, resulting in reduced inconsistencies and enhanced signal-to-noise ratio (SNR). Subsequently, the corrected image data is used for 3Dreconstruction. Although simulation based artefact correction is capable of enhancing 3D image quality, physically correct simulation of projection data can be challenging in terms of accuracy and runtime. Especially when the simulation includes a large variety of artefacts. The aim of this paper is to show the potentials and limitations of simulation based artefact correction algorithms and to discuss the need for full Monte-Carlo method. Furthermore, the requirements for optimal simulation results are determined