Warping Nodal Levels Via M-Rep Based Diffeomorphism
Rohit R. Saboo, Sarang C. Joshi, Joshua H. Levy, Stephen M. Pizer, J. Rosenman · 2006
The identification of lymph node drainage levels in the neck from CT scans is necessary for the proper treatment of cancer of the head and neck with radiation therapy. However, to comply with the guidelines one must segment at least 20 fiducial anatomic objects on each side, and then apply complex rules to determine six boundaries for each of nine different levels. To reduce physician time and standardize results it is necessary to automate the entire process. In this paper, we propose solving this problem by deforming the nodal levels from an atlas patient to target patient in a way combining two novel ideas: a) The large deformation diffeomorphic transformations of curved landmark paths; b) a scheme provided by m-reps and their geodesic paths for producing curved paths between corresponding points associated with the fiducial objects. This method is applicable to a wide variety of problems of mapping between an atlas and a target image with automatically segmentable structures. 1.