Using wavelet transform and neural networks for the analysis of brain MR images

In this study brain MR images are segmented into the constitutive tissues such as the gray matter, white matter and cerebrospinal fluid using multiresolutional wavelet packet transform and self-organizing map networks. For this purpose T1-weighted, T2-weighted and PD-weighted simulated brain MR images are used. First of all, wavelet packet transform is applied to the images. Subimages obtained from the transform are filtered using best subtree method. Feature vector that is used as input to the neural network is constructed by combining the reconstructed images that are the result of the transform. As a consequence brain MR images are segmented into gray matter, white matter and cerebrospinal fluid using self-organizing map networks.