A0627
Title: Statistical analysis of locally parameterized shapes
Authors: Mohsen Taheri Shalmani - University i Stavanger (Norway) [presenting]
Joern Schulz - University of Stavanger (Norway)
Abstract: The establishment of correspondence and defining shape representation are crucial steps in statistical shape analysis for detecting local dissimilarities between two groups of objects. Most shape representations are based on either noninvariant spatial properties to rigid transformation or extrinsic geometric properties. Shape analysis based on extrinsic geometric properties could be misleading, and based on noninvariant properties is biased because the act of alignment is necessary. Also, mathematical interpretation of the type of dissimilarity, e.g., bending, elongation, twisting, protrusion, etc., is desirable. By defining local coordinate systems on object skeletal, a novel shape representation based on intrinsic and invariant object properties will be discussed. The proposed shape representation is also superior for simulation and skeletal deformation. The power of the hypothesis testing based on the introduced shape representation is demonstrated in the simulated data as well as the left hippocampi of patients with Parkinson's disease versus a healthy control group.
