Andreas Krauss PhD, Abstract Co-Author: Employee, Siemens AG

Rainer Raupach PhD, Abstract Co-Author: Employee, Siemens AG

Bernhard Schmidt PhD, Abstract Co-Author: Employee, Siemens AG

Thomas G. Flohr PhD, Presenter: Employee, Siemens AG

To quantitatively evaluate the performance of a novel algorithm for metal artifact reduction in computed tomography (CT).

The proposed iterative metal artifact reduction algorithm starts with standard CT reconstruction. Metal pixels are then identified through HU thresholding and a metal sinogram is generated by forward projection. A prior image is calculated from the initial image by assigning soft tissue pixels (identified by upper and lower HU thresholds) to 0 HU. Then normalized sinogram interpolation is performed: The prior image is forward projected and the original rawdata is normalized pixelwise with the prior sinogram. In the normalized sinogram, pixels within the metal trace are replaced by linear interpolation from the edges of the metal trace. The interpolated sinogram is de-normalized and standard reconstruction of the corrected sinogram is performed. The procedure is repeated iteratively with the output of the previous iteration used as input for prior image calculation. Finally, a frequency split is performed to preserve valid edge information of the non-corrected images: The high frequency part of non-corrected images is merged with the low frequency part of MAR-corrected images. The algorithm was applied to 10 hip replacement cases.  

Streak artifacts from hip prostheses as well as the typical dark band between bilateral hip prostheses were almost completely eliminated. Pelvic soft tissue and organ structure was restored. Typical drawbacks of previous MAR algorithms such as introduction of new artifacts or compromised bone structure close to the prostheses were minimal due to iterative normalized interpolation and frequency split, respectively. Between bilateral hip prostheses, mean HU values within regions of interest located inside the bladder (expected to be water, i.e. 0 HU) were improved from -440 HU to -7 HU on average; the maximum improvement was from (-732 ± 252) HU to (-7 ± 38) HU.

The proposed algorithm substantially reduces artifacts from hip prostheses.

The proposed algorithm strongly facilitates the visualization of pelvic anatomy. Due to the recovery of HU values, improvements for radiotherapy treatment planning are expected.

Krauss, A, Raupach, R, Schmidt, B, Flohr, T, Iterative Metal Artifact Reduction in Computed Tomography.  Radiological Society of North America 2013 Scientific Assembly and Annual Meeting, December 1 - December 6, 2013 ,Chicago IL. http://archive.rsna.org/2013/13019575.html