Quantitative Comparison of Registration-Based Lung Motion Estimates from Whole-Lung MR Images and Corresponding Two-Dimensional Slices

Introduction: Quantification of pulmonary deformation is useful in characterizing normal lung motion as well as the changes that occur as a result of pathological processes [4]. Magnetic resonance (MR) imaging of the lung can be used to noninvasively visualize and assess pulmonary anatomy and dynamics [2,3]. Ideally, motion quantification would be performed using whole-lung images to eliminate the confounding effect of through-plane motion. Currently, real-time imaging of deforming lungs is more easily performed at a single slice position. Here, we seek to quantify how much information about pulmonary kinematics is lost when estimating whole-lung motion at single two-dimensional (2-D) slice positions. We compare lung motion estimated via image registration from whole-lung datasets with the motion within a selected set of slices extracted from the whole-lung images. Our hypothesis is that the accuracy of lung motion measured from a single slice is affected by both the choice of imaging plane and the degree of deformation within the lung (estimated by the percent of instantaneous lung volume expired) during the specified interval of respiration.