3D Object Classification and Segmentation Methods - High-Quality Visual Experience - page 343

Fig. 4 Segmentation and classification procedure
In classification step, after removing parts of no interest, aim is to isolate the
brain mass and segment it into grey matter, white matter and cerebro-spinal fluid. By
assigning each voxel to an intensity class and each intensity class to a tissue class, all
the voxels of the 3D data set can be attributed to a tissue class and local shapes can
be estimated. It first extracts a set of features from the input database and then de-
signs a neural network classifier for the specific detection task using those features.
Results of brain segmentation and classification are shown in Fig. 6.