Iterative geometric calibration in circular cone-beam computed tomography

The reconstructed image may be distorted by geometry artifacts due to the presence of mechanical misalignments in circular cone-beam computed tomography (CT) system. To avoid geometry artifacts intervention, a novel geometric calibration method belonging to the iterative-type methods, which uses a dedicated calibration phantom to determine all geometry parameters, is presented in this paper. The phantom consists of two steel balls, and the trajectories of two ball centers are two circles in circular scan. Minimization of the errors between the predicted trajectories of ball center and the ideal circles is the principle in the proposed method. In summary, the proposed method consists of three parts. Firstly, an efficient description of geometry which reduces the number of geometry parameters from 7 to 4 is stated. Then, a nonlinear least square model depending on the 4 parameters is derived. At last, the Levenberg–Marquardt algorithm (LMA) is applied to solve the optimization problem. We implemented the proposed method and the classical analytic method presented by Noo et al. on real data in our laboratory cone-beam CT. The results indicated that our method could provide satisfactory reconstructed images as good as Noo's method.