Estimation of Lung Motion Using Deformable Image Registration Based on Compressible Flow

Accurate estimation of respiratory motion is of great significance in many medical applications because respiratory motion introduces difficulty in planning and delivering radiotherapy treatment to tumors in the thorax and upper abdomen. In the present investigation, an algorithm for estimation of lung motion in four-dimensional computed tomography (4DCT) images has been proposed. Specifically, spatial correspondence is established between phase images in terms of displacement field which is initially generated by compressible flow based method and then refined by displacement regularization. By linking displacement field at discrete phases, we obtain the time-related continuous displacement field which allows for locating any lung point at any time in respiratory cycle. We Apply the proposed method to 5 patients' 4DCT data and use a large number of landmark point pairs to evaluate the performance, the average spatial error +/- standard deviation is reduced from 6.71 +/- 4.47 mm to 1.20 +/- 0.78 mm. Results indicate that the proposed method achieves significantly improved accuracy in the estimation of lung motion.