MIAL

Talk by Prof. Klaus Toennies: Dynamically deformable models for shape representation

January 22, 2007
Location: TASC 9204 West, SFU
Presenter: Prof. Klaus Toennies (Medical Image Analysis)

Abstract
Dynamically deformable models (DDMs) have been used for shape synthesis in order to carry out simple shape based analysis tasks. A DDM represents an average 2-d or 3-d shape of an object class. Shape variations are constrained by internal forces similar to those of an active contour or active shape. Local shape constraints constrain global shape variation. A model instance placed into an image simulates a physical entity and can search an object in a physically interpreted image data set. Two applications of for a DDM will be presented in the talk: For registering an anatomic model of the left ventricle of the heart to functional SPECT, we developed a 3-d DDM on the basis of a mass spring model with an additional shape-preserving constraint. The constraint separates size variation from shape variation which is important because the former is acceptable for accommodating different heart shape while the latter is not. Application of the DDM enabled fast and fully automatic registration of the model with the data. In another application searching the Heschl gyrus in flatmaps of the cortical surface from MRI, we developed a hierarchical 2-d DDM based on FEMs that supports a simple recognition-by-parts strategy in order to find the U-shaped gyrus among a large number of such shapes. Recognition is not fully automatic but provides the user with a number of alternatives among which he has to select.

Short Biography of Presenter

1987, PhD in Computer Science, Techn. Univ. Berlin
1987-89, Res. Asst. Prof., Medical Image Processing Group, University of Pennsylvania
1989-1995, Asst Prof, Computer Graphics Group, Techn Univ Berlin
1996-1998, Head, Image Processing Group, Dept Radiology, Univ Berne, Switzerland
1998-present, Prof, Image Proc and Understanding Group, Univ Magdeburg (3 months in 2002): Research Fellow UBC and SFU


Site by Faculty of Applied Science web team, Simon Fraser University