Fast method for computing a system matrix using a polar-coordinate pixel model with concentric annuluses of different radial widths.

Despite better reconstruction quality for incomplete or noisy projection data compared to analytic reconstruction, computed tomography iterative techniques are time-consuming, mainly due to high system matrix computation. A polar-coordinate pixel model with concentric annuluses of different radial widths was established and a fast method for computing the system matrix was presented based on characteristics of this model. Compared with the Siddon algorithm and an efficient Cartesian algorithm introduced by Zhang, the proposed algorithm based on the simultaneous algebraic reconstruction technique shows speed advantages for both numerical simulation and experiment, without noticeable loss of image quality.