Artefact Reduction Methods for Iterative Reconstruction in Full-fan Cone Beam CT Radiotherapy Applications

A cone beam CT (CBCT) system acquires two-dimensional projection images of an imaging object from multiple angles in one single rotation and reconstructs the object geometry in three dimensions for volumetric visualization. It is mounted on most modern linear accelerators and is routinely used in radiotherapy to verify patient positioning, monitor patient contour changes throughout the course of treatment, and enable adaptive radiotherapy planning. Iterative image reconstruction algorithms use mathematical methods to iteratively solve the reconstruction problem. Iterative algorithms have demonstrated improvement in image quality and / or reduction in imaging dose over traditional filtered back-projection (FBP) methods. However, despite the advancement in computer technology and growing availability of open-source iterative algorithms, clinical implementation of iterative CBCT has been limited. This thesis does not report development of codes for new iterative image reconstruction algorithms. It focuses on bridging the gap between the algorithm and its implementation by addressing artefacts that are the results of imperfections from the raw projections and from the imaging geometry. Such artefacts can severely degrade image quality and cannot be removed by iterative algorithms alone. Practical solutions to solving these artefacts will be presented and this in turn will better enable clinical implementation of iterative CBCT reconstruction.