Geometry calibration method for a cone‐beam CT system

Purpose The positioning accuracy of each component is important to ensure the image quality of cone‐beam CT . However, accurate positioning is not easy and requires experience and time. The option is to calibrate the geometric parameters and then plug them into a reconstruction algorithm which is the preferred solution in practice. In this case, the image quality is determined by the accuracy and precision of the calibration method. This work describes a method to independently calibrate an imaging system in each pose (projection angle) for a cone‐beam CT with a nonideal circular trajectory. Methods The calibration method uses a phantom with 12 beads on 2 planes that are observed on the radiographic images. This pose‐independent calibration method ( PIC ) can decorrelate the relationships among the geometric parameters so that the parameters can be estimated one‐by‐one. This simplifies the calibration process. Besides the pose‐independent calibration method, this paper also describes an extended calibration method with additional constraints on the system geometry. Both methods are validated with numerical simulations and then experimentally on a practical system with a scanning object loosely supported by rotating wheels. The object rotates during the CT data acquisition. The angular and pose information of the CT system are not accurately known a priori in this case. Results The numerical simulations and the experiments both provide satisfactory results. The relative error of the calibrated source‐to‐detector distance in the simulation is less than 0.1%. The errors in the calibrated roll, pitch, and yaw angles are less than 0.04°. A sensitivity study using various bead position uncertainties in random directions shows that the pose‐independent calibration method is robust to measurement errors. Tests were also done with a nonideal circular trajectory for further validation. Images reconstructed using the geometric parameters from both the pose‐independent and the extended calibration methods were free of artifacts and blur from misalignment. All these results demonstrate the effectiveness of these methods. Conclusions The PIC method can be used to independently calibrate the geometric parameters of a cone‐beam CT view‐by‐view. Thus, the PIC method can be implemented on commonly used systems such as circular, nonideal circular or C‐arm cone‐beam CT s. The PIC method can also be useful for some irregularly configured CT systems to fulfill special imaging requirements, for example, a CT system when the x ray source or the rotating platform cannot be easily located. The PIC method will reduce the costs of ensuring very precise mechanics and the labor in fine tuning CT systems.