Proceedings Article | 6 March 2013
Proc. SPIE. 8668, Medical Imaging 2013: Physics of Medical Imaging
KEYWORDS: X-ray computed tomography, CT reconstruction, Imaging systems, Sensors, Metals, Image segmentation, X-rays, 3D modeling, Signal processing, Reconstruction algorithms
In Cone Beam CT Imaging, metallic and other dense objects, such as implantable orthopedic appliances, surgical clips
and staples, and dental fillings, are often acquired as part of the image dataset. These high-density, high atomic mass
objects attenuate X-rays in the diagnostic energy range much more strongly than soft tissue or bony structures, resulting
in photon starvation at the detector. In addition, signal behind the metal objects suffer from increased quantum noise,
scattered radiation, and beam hardening. All of these effects combine to create nonlinearities which are further amplified
by the reconstruction algorithm, such as conventional filtered back-projection (FBP), producing strong artifacts in the
form of streaking. They reduce image quality by masking soft tissue structures, not only in the immediate vicinity of the
dense object, but also throughout the entire image volume. A novel, physical-model-based, metal-artifact reduction
scheme (MAR) is proposed to mitigate the metal-induced artifacts. The metal objects are segmented in the projection
domain, and a physical model based method is adopted to fill in the segmented area. The FDK1 reconstruction algorithm
is then used for the final reconstruction.
