Modeling and estimation of detector response and focal spot profile for high-resolution iterative CT reconstruction

High spatial resolution in x-ray computed tomography (CT) is paramount in many clinical applications. The realization of high-resolution CT requires both high-precision imaging hardware and well-designed image reconstruction algorithms. Compared to the conventional filtered backprojection (FBP), one attractive advantage of model-based iterative reconstruction (MBIR) is the ability to incorporate more realistic models of the blurring effects caused by the detector response and the focal spot. Several studies have shown that modeling such effects in image reconstruction improves spatial resolution and reduces artifacts. However, applying such advanced forward models to real CT data remains challenging, because the system response of a real scanner is seldom known exactly, and improper choice of modeling parameters may introduce reconstruction artifacts. Here we report the results of finite beam-width modeling for a GE Discovery CT 750 HD scanner operating in high resolution mode using focal spot wobble. We introduced modeling of gaps between detector cells and non-uniform focal spot profile, and evaluated how various combinations of modeling parameters affect the reconstructed images. We showed that by properly adjusting the model parameters, in-plane spatial resolution was improved from 15 to 17 lp/cm, as visualized with a Catphan resolution gauge. We further propose a maximum-likelihood approach to estimate the detector response function and the focal spot profile along with the CT image, which may be used to further refine the forward model.