Application of artificial neural networks for tissue classification from multispectral magnetic resonance images of the head

Abstract

The suitability of artificial neural networks (ANNs) for the classification of multispectral magnetic resonance images (MSMRI) is explored. MSMRI feature space distributions of various tissues and phantoms were examined to determine if the data is more suitable for ANN classification, as opposed to classical Bayesian approaches for intensity-based classification. Additionally, MSMRI normalization methods were investigated to determine suitability for improving feature space distributions independent of classification methods.<>