ANATOMICAL OBJECT RECOGNITION USING DEFORMABLE GEOMETRIC-MODELS

This work addresses the problem of identifying specific objects within three-dimensional data sets. In the specific example chosen we are tr ying to locate part of the brainstem, and associated structures of the upper spinal cord and mesencephalon, and determine its size, shape an d orientation. The approach is in two parts: firstly, to use a control led. deformable model, based on superquadric geometric primitives as a n initial estimate and apply genetic algorithms as the technique for s olving the complex optimization problem of defining an approximate enc ompassing envelope within which the object will be found. The second s tep implements a segmentation technique, based on image features, to r efine the tentative object into a more complex, and realistic, shape s uitable for subsequent visualization or volumetric measurement.