2007 Conference Objects

Symmetry Based Multi-modality Registration of the Brain Imagery

Liu, Xin; Imielinska, Celina Z.; Laine, Andrew F.; D'Ambrosio, Anthony Louis

We propose a framework for multi-modality brain registration methods using symmetry plane as the principal feature for geometric matching. By bringing the symmetry plane of two rigid objects into coincidence, we can potentially match two objects approximately ifthere is no apparent elastic distortion. We illustrate this concept using Visible Human dataset, that includes color cryosection, and radiological data, from which we extract 3D mesh models of skin, brain and skull, and aligned them into nested bodies. Each model was generated with different spatial orientation and resolution, and their alignment, was guided by the underlying anatomical relationships in the head region, and a requirement, by an application that the meshes didn't intersect. After alignment of the symmetry planes, obtained for each mesh using spatial affine transformations, the further geometric adjustment, to achieve complete registration of the nested models, is confined within a 2D plane (i.e. the symmetry plane). This simple method of registration of mesh anatomical models, has a potential to significantly reduce the degrees of freedom in various 3D brain registration applications. It can be also treated as a pre-registration operation before applying other registration methods.