Comparison Between Projection Weighting and Rebinning Approaches in Analytical Image Reconstruction in Off-Center Flat Panel Cone-Beam CT Imaging

Abstract

In cone-beam CT (CBCT) imaging, a half-fan (HF) mode acquisition can be used to obtain a larger field-of-view (FOV) without physically increasing the detector size. However, this leads to an asymmetric sinogram with a truncated set of projection that needs to be considered during image reconstruction., especially when using analytical methods such as the filtered backprojection (FBP) and its 3D approximation, the Feldkamp-Davis-Kress (FDK) algorithm. Various approaches have been suggested in literature, which can be roughly categorized as weighting based and projection rebinning approaches. In this paper, a comparison between the two approaches are made using the XCAT phantom, based on the geometry of the Varian On-Board Imager (OBI) as the CBCT acquisition device that is usually used for image guided radiotherapy (IGRT) during treatment delivery. It is found that analytical reconstruction alone will produce a percent normalized root mean squared error (NRMSE) of 36.81% in the reconstructed image, whereas using the weighting based and rebinning approaches reduces this error to 22.86% and 4.68% respectively. Apart from that, the weighting based approach is also found to be on average biased and produces an overall lower intensity in the reconstructed image whereas the rebinning approach does not have this problem, as well as having a smaller standard deviation of the intensity deviations from the original phantom image as the ground truth.