Multi-featured Multi-scale Deformable Model for Three-dimensional Surface Extractio from Volumetric Image

Deformable surface models provide a very attractive method for segmenting the three-dimensional shape of complex and various internal organs from volumetric medical images. However, although researchers have succeeded in overcoming some of the limitations of this approach, there still exist several significant unsolved problems. In this paper, we propose a multi-featured multi-scale deformable surface model as a solution to three of these problems: model initialization dependency, inability to extract object concavities, and model selfintersection. The multi-scale approach, in which w progressively resample our deformable triangulated surface model globally and locally in order to match its resolution to the level of a 3-D image pyramid, provides insensitivity to local minima and model initialization as well as the ability to extract object concavities . The multifeatured energy function including internal, external, and geometric self-collision constraint force prevents selfintersection effectively. We have applied our model to the challenging problem of brain cortex boundary extraction and preliminary results are presented.