PURPOSE: To evaluate deformable mesh registration (DMR) as a tool for validating automatic target registration algorithms used during image-guided radiation therapy. METHODS: DMR was implemented in a hierarchical model, with rigid, affine, and B-spline transforms optimized in succession to register a pair of surface meshes. The gross tumor volumes (primary tumor and involved lymph nodes) were contoured by a physician on weekly CT scans in a cohort of lung cancer patients and converted to surface meshes. The meshes from weekly CT images were registered to the mesh from the planning CT, and the resulting registered meshes were compared with the delineated surfaces. Known deformations were also applied to the meshes, followed by mesh registration to recover the known deformation. Mesh registration accuracy was assessed at the mesh surface by computing the symmetric surface distance (SSD) between vertices of each registered mesh pair. Mesh registration quality in regions within 5 mm of the mesh surface was evaluated with respect to a high quality deformable image registration. RESULTS: For 18 patients presenting with a total of 19 primary lung tumors and 24 lymph node targets, the SSD averaged 1.3 ± 0.5 and 0.8 ± 0.2 mm, respectively. Vertex registration errors (VRE) relative to the applied known deformation were 0.8 ± 0.7 and 0.2 ± 0.3 mm for the primary tumor and lymph nodes, respectively. Inside the mesh surface, corresponding average VRE ranged from 0.6 to 0.9 and 0.2 to 0.9 mm, respectively. Outside the mesh surface, average VRE ranged from 0.7 to 1.8 and 0.2 to 1.4 mm. The magnitude of errors generally increased with increasing distance away from the mesh. CONCLUSIONS: Provided that delineated surfaces are available, deformable mesh registration is an accurate and reliable method for obtaining a reference registration to validate automatic target registration algorithms for image-guided radiation therapy, specifically in regions on or near the target surfaces.
PURPOSE: To evaluate deformable mesh registration (DMR) as a tool for validating automatic target registration algorithms used during image-guided radiation therapy. METHODS:DMR was implemented in a hierarchical model, with rigid, affine, and B-spline transforms optimized in succession to register a pair of surface meshes. The gross tumor volumes (primary tumor and involved lymph nodes) were contoured by a physician on weekly CT scans in a cohort of lung cancerpatients and converted to surface meshes. The meshes from weekly CT images were registered to the mesh from the planning CT, and the resulting registered meshes were compared with the delineated surfaces. Known deformations were also applied to the meshes, followed by mesh registration to recover the known deformation. Mesh registration accuracy was assessed at the mesh surface by computing the symmetric surface distance (SSD) between vertices of each registered mesh pair. Mesh registration quality in regions within 5 mm of the mesh surface was evaluated with respect to a high quality deformable image registration. RESULTS: For 18 patients presenting with a total of 19 primary lung tumors and 24 lymph node targets, the SSD averaged 1.3 ± 0.5 and 0.8 ± 0.2 mm, respectively. Vertex registration errors (VRE) relative to the applied known deformation were 0.8 ± 0.7 and 0.2 ± 0.3 mm for the primary tumor and lymph nodes, respectively. Inside the mesh surface, corresponding average VRE ranged from 0.6 to 0.9 and 0.2 to 0.9 mm, respectively. Outside the mesh surface, average VRE ranged from 0.7 to 1.8 and 0.2 to 1.4 mm. The magnitude of errors generally increased with increasing distance away from the mesh. CONCLUSIONS: Provided that delineated surfaces are available, deformable mesh registration is an accurate and reliable method for obtaining a reference registration to validate automatic target registration algorithms for image-guided radiation therapy, specifically in regions on or near the target surfaces.
Authors: Esben S Worm; Anders T Hansen; Jørgen B Petersen; Ludvig P Muren; Lars H Præstegaard; Morten Høyer Journal: Acta Oncol Date: 2010-10 Impact factor: 4.089
Authors: Elisabeth Weiss; Scott P Robertson; Nitai Mukhopadhyay; Geoffrey D Hugo Journal: Int J Radiat Oncol Biol Phys Date: 2011-12-22 Impact factor: 7.038
Authors: Gerald Lim; Andrea Bezjak; Jane Higgins; Doug Moseley; Andrew J Hope; Alex Sun; John B C Cho; Anthony M Brade; Clement Ma; Jean-Pierre Bissonnette Journal: J Thorac Oncol Date: 2011-03 Impact factor: 15.609