Metal Artefact Reduction Framework in Dental Cone-Beam CT (CBCT)

Objectives:
Highly dense object cause well-known streak artefacts in CBCT. Working on the projection images used for 3D-reconstruction, a metal artefact reduction framework is presented to reduce these streak artefacts by means of integration of surrounding tissue information.
Methods:
Highly dense artefact sources (dental implants, fillings or metal restoration) are detected in all x-ray projection radiographs used for 3D-reconstruction. The regions are either ignored or replaced by surrounding information (inpainting) and the resulting projection images are used within the 3D reconstruction process. Two methods are being compared: an appropriate inpainting method versus a missing value implementation. GPU-based iterative 3D reconstruction using algebraic reconstruction techniques (ART) is employed for this purpose and the algorithms are tested on simulated data (extended Shepp-Logan phantom) and real-world CBCT data.
Results:
We observe significantly improved reconstructions in the vicinity of the artefact sources when compared to non-corrected original images. Although we omit the artefact sources from the reconstruction process, their 3D positions are computable and supplement the 3D volume visualization.
Conclusions:
Our metal artefact reduction strategies concentrate on ignoring or replacing image values in the projection images because the non-linear attenuated projection data are useless for reconstruction in many cases. Even for various independent artefact sources we obtain sufficient results in comparison to present reconstruction algorithms used in dental CBCTs.