In the previous section, we explained the underlying theory and general properties of the GRBF neural networks used in this chapter. This provides the basis for their application to automatic segmentation of magnetic resonance imaging data sets of the human brain.
The general concept of multispectral voxel-based brain segmentation can be explained as follows: n different 3D data sets for each brain are obtained employing different MRI acquisition parameters. In this chapter, we used n = 4 MRI acquisition sequences (T1 weighted, T2 weighted, proton density weighted, and inversion recovery sequences; see Section 5). Segmentation aims at classifying each voxel of the multi-spectral data set as belonging to a specific tissue type, thus obtaining information about sturucture and volume of the tissue classes.
A classical problem with numerous clinical applications is the segmentation of brain imaging data with respect to the tissue classes gray matter, white matter, and cerebrospinal fluid (CSF). Several other structures such as meninges or venous blood may be introduced as additional segmentation classes. However, these additional classes comprise only a small part of the total brain volume. Furthermore, for most of the clinical applications, the focus of interest is reduced to gray- and white-matter structures. Therefore, we assigned these minor additional classes to CSF.
Although such a threefold classification of brain tissue may be sufficient for numerous clinical applications, it should be emphasized that the concept presented in this chapter can be extended to an arbitrary number of tissue classes. Especially one may think of introducing additional classes for the identification of pathological tissue, e.g., multiple sclerosis plaques or malignant brain tumor structures.
TABLE 1 Acquisition parameters for the four MRI sequences of the 3D data sets
Was this article helpful?