A New Hybrid Segmentation Method for MR Images

Wang Song,Liang Yilong,Shaowei Xia,Zesheng Tang,Wenjing Huang
1998-01-01
Abstract:This paper introduces a hybrid segmentation method for 3D MR image of the head. The hierarchical architecture is adopted in this method. At first, the adaptive threshold algorithm is used to extract the head surface and the brain surface. Then we use Guassian filters to gain the multiscale features of the brain voxels. To reduce the influence of intensity inhomogeneity between and within the 2D MR slices, a new classifier combining the unsupervised clustering and supervised feed-forward neural network (NN) is developed to label the brain voxels into gray matter and white matter. Through the training of feed-forward NN with back-propagation (BP) algorithm, the expert knowledge can be included. Both the comparison results and 3D visualization show the performance of this new segmentation method.
What problem does this paper attempt to address?