Marcel Beister, Presenter: Employee, CT Imaging GmbH

Daniel Kolditz, Abstract Co-Author: Employee, CT Imaging GmbH

Willi A. Kalender PhD, Abstract Co-Author: Consultant, Siemens AG Consultant, Bayer AG Founder, CT Imaging GmbH Scientific Advisor, CT Imaging GmbH CEO, CT Imaging GmbH

Model-based iterative reconstruction (MBIR) methods model the CT acquisition process more accurately than conventional techniques such as filtered backprojection (FBP). In addition to reducing dose levels, a reduction of image artifacts due to, e.g., beam hardening and sampling errors, and of noise are feasible. We investigated the potential of MBIR to reduce artifacts caused by metallic implants using an extended algorithm which includes physical modeling of the polychromatic photon spectrum.

We implemented MBIR including photon statistics modeling, geometrical modeling of the CT system, modeling of a realistic local image-neighborhood and a novel method for spectral modeling. By segmenting the current volume estimate during the iteration process into different materials and considering multiple energy bins of the x-ray spectrum, the forward projection can be done in a polychromatic way. Thus modeling of beam hardening effects becomes an integral part of the reconstruction process. The segmentation can be improved using prior knowledge on e.g. the material and shape of the metallic object. The algorithm was accelerated with graphical processing units using CUDA (NVIDIA, Santa Clara, USA). Evaluation included simulated and measured datasets of various CT systems. We compared the results of MBIR to FBP with and without interpolation-based metal artifact reduction (MAR). Image noise, CT value accuracy, and low-contrast detectability were analyzed using different phantoms with and without metallic inserts.

Besides lower image noise, the extended MBIR approach provided strong reduction of beam hardening artifacts through the modeling of the polychromatic acquisition process. The typical streak artifacts oriented tangentially to metallic objects were reduced strongly by geometric modeling of the acquisition system. CT value accuracy, detectability of soft-tissue as well as visibility of structures surrounding metallic implants were markedly improved by the proposed method.

The proposed model-based reconstruction provided image quality superior to both uncorrected and MAR-corrected FBP especially if further high-contrast structures are present in the image.

The use of model-based reconstruction including spectral modeling reduces various artifacts caused by metallic objects and can thereby provide improved diagnostic information.

Beister, M, Kolditz, D, Kalender, W, Metal-Artifact Reduction in Computed Tomography Using Model-based Iterative Reconstruction with Spectral Modeling.  Radiological Society of North America 2012 Scientific Assembly and Annual Meeting, November 25 - November 30, 2012 ,Chicago IL. http://archive.rsna.org/2012/12043928.html