A deformable neuroanatomy textbook based on viscous uid mechanics

This paper demonstrates a novel image recognition method based on Grenanders Global Shape Models. A mathematical textbook is constructed to represent the typical structure of a human brain. Next, a set of probabilistic transformation based on the theory of viscous uids is applied to the textbook to generate a rich family of brain images. We describe a method that estimates the maximum a posteriori transformation which deforms the coordinate system of the textbook into the coordinate system of a particular individual. Once this transformation is determined, information about the individual brain can be queried by applying the transformation to the symbolic data contained in the textbook.