Employing shape to aid magnetic resonance brain segmentation

As part of an ongoing project for automatic segmentation of magnetic resonance (MR) brain images, the use of a psychologically and neuropsychologically motivated representation of shape is explored. Shapes are recognized and learned in a translation and scale invariant manner, and then are used to aid segmentation. Object recognition is achieved after training the system to associate observed shapes and their relative locations with an object defined by a human expert. The inspiration and current implementation of a shape recognition module are described, together with some preliminary results