Investigation of X-Ray Intensity-Linear Models in CT Image Reconstruction

Abstract

Metal artifacts remain a significant challenge in X-ray computed tomography (CT) image reconstruction. In this study, we explore a mathematical model–based approach aimed at more accurately representing the underlying physical phenomena. Specifically, we investigate an image reconstruction method that employs a cost function linear with respect to X-ray intensity, in contrast to the conventional logarithmic-intensity–based cost function used in iterative reconstruction algorithms. Experiments were conducted using a three-dimensional cone-beam projection CT system with both simulated and real projection data containing metallic objects. The proposed approach demonstrated a reduction in typical metal artifact patterns compared with conventional methods. However, developing a fast optimization algorithm with guaranteed convergence remains an open problem.