Respiratory motion estimation from slowly rotating X-ray projections

As radiotherapy has become increasingly conformal, geometric uncertainties caused by breathing and organ motion have become an important issue. Accurate motion estimates may lead to improved treatment planning and dose calculation in radiation therapy. However, respiratory motion is difficult to study by conventional X-ray CT imaging since object motion causes inconsistent projection views leading to artifacts in reconstructed images. We propose to estimate the parameters of a nonrigid motion model from a set of projection views of the thorax that are acquired using a slowly rotating cone-beam CT scanner, such as a radiotherapy simulator. We use a conventionally reconstructed 3D thorax image, acquired by breath-hold CT, as a reference volume. We represent respiratory motion using a flexible parametric nonrigid motion model based on B-splines. The motion parameters are estimated by optimizing a regularized cost function that includes the squared error between the measured projection views and the reprojections of the deformed reference image. Preliminary 2D simulation results show that there is good agreement between the estimated motion and the true motion.