A Deformable Statistical Shape Model Applied to Three-Dimensional Lumbar Vertebra Images

A surface morphing technique utilizing the Generalized Gradient Vector Flow (GGVF) method was implemented and tested. The surface morphing results in a homeomorphic deformation of a reference surface onto individual members of a data set. Internal and external forces that drive the surface morphing result in correspondences between the parameterized surfaces. Principal component analysis was used to form a statistical shape model from the resulting surfaces. Three dimensional L1-lumbar vertebral images were used to demonstrate the algorithm. Volume image preprocessing and segmentation is described as well.