Seminars

Although shape asymmetry has been investigated in many branches of science, there is a lack of unified methodological framework for quantifying local shape asymmetry. Previous literature mainly deals with quantifying a global amount of shape asymmetry. A more interesting question would be to ask if we could spatially localize the source of asymmetry.

In brain imaging, this question has been successfully addressed by using the deformable template approach of Grenander and Miller. By registering the original and its mirror reflected 3D magnetic resonance image (MRI), one can establish the correspondence across hemispheres and, in turn, able to construct the localize asymmetry index of type (L-R)/(L+R). The additional computational burden of establishing deformation across hemispheres and possible mismatching of sulcal pattern across subjects are two major shortcomings of this widely used approach.

In this talk, we present a different framework for shape asymmetry analysis that basically combines the deformable template idea and Brechbuler's 3D Fourier descriptor. Surface shape registration, surface data smoothing and surface parameterizations are all tackled in a unified framework. This is a joint work with Kim Dalton and Richard Davidson of the Waisman Laboratory for Brain Imaging and Behavior. An application of the same technique to longitudinal mandible shape modeling (in collaboration with Houri Vorperian of the Vocal Tract Development Laboratory) on 300 subjects will be also
briefly discussed.