Purpose: To reconstruct the motion manifold defined on anatomical interfaces from real‐time MR imaging. Methods∷We have developed a novel method to construct the evolving motion manifold in real‐time by preserving the point correspondence when a level set function is updated to track moving anatomical interfaces. The key to our development is to define a correspondence function in a Eulerian system and advecting it along the level set function under the same velocity. The motion manifold is ‘grown’ by concatenating the point correspondence maps. The proposed method was tested with renal MR image sequences under both synthetic and physiological motion. The synthetic sequence was generated by shifting a static MR image continuously in one dimension to a maximum of 29mm. Performance of the estimation was analyzed by comparing the reconstructed motion manifold and the known truth. Physiological motion manifold was reconstructed from real‐time EPI data acquired from human subjects under heavy breathing condition. Results: In the synthetic motion test, the reconstructed motion manifold translates each particle to closely follow the underlying motion. The mean error along and perpendicular to the motion direction is −2.58mm and 0.03mm respectively. These are reasonable results considering the low signal‐to‐noise ratio of the real‐time MR images with pixel size equals 1.52mm. The motion manifold reconstructed from physiological real‐time renal MR exhibits a semiperiodic pattern that agrees with respiratory behavior. Spatially varying motion pattern from different parts of the kidney interface can be observed. Conclusions: We have proposed a novel level set method to track and estimate anatomical moving interface while preserving the point correspondence during the dynamic evolution. The generated motion manifold provides important knowledge about the characteristics of a particular organ movement and facilitates further model development for retrospective analyzing dose impact of motion or predicting motion for real‐time delivery adaptation. NIH grant 1R01CA159471. PI: Michael Gach.
|Number of pages||1|
|State||Published - Jun 2012|