Patient motion correction in computed tomography by reconstruction on a moving grid

This work describes a method of motion correction in image reconstruction for both emission computed tomography and X-ray CT. The method assumes that moving objects are represented as objects of constant intensity defined on a deformable grid. Large-scale body motion and respiratory motion are tracked by fiducial markers attached to the patient's body. The positions of the markers are determined using a camcorder-based system. The deformation of the body is approximated by local shifts in the grid nodes of the image representation, which are linear functions of the time-dependent displacement of the fiducial markers. After the grid motion is established, node intensities are reconstructed using either an analytical or an iterative statistical reconstruction algorithm. The information about the grid deformation is incorporated during the computation of the forward projection. Both analytical and iterative ML-EM algorithms are inplemented without a significant increase in reconstruction time. Numerical experiments are presented in a two-dimensional case for a simulated CT scan of a deformable phantom and in a three-dimensional case for the NCAT phantom with respiratory motion.