Optimal deformable surface models for 3D medical image analysis.

摘要: We present a novel method for building an optimal statistical deformable model from set of surfaces whose topological realization is homeomorphic to compact 2D manifold with boundary. The parameterization each shape recursively refined by using hierarchical PBMs and tensor product B-spline representation the surface. A criterion based on MDL used define internal correspondence training data. strength proposed technique demonstrated deriving concise human left ventricle which has principal modes variation that correspond intrinsic cardiac motions. demonstrate how derived can be 3D dynamic volume segmentation ventricle, its accuracy assessed comparing results obtained manual delineation cine MR data 8 asymptomatic subjects. extension shapes complex topology also discussed.