Possible changes in brain metabolites are associated with Motor Neurone Disease (MND). Magnetic Resonance Spectroscopy (MRS) has been performed on the brains of MND patients and control volunteers to acquire signals which contain information about brain metabolites from within the motor cortex area. Discrimination between JvThD and normal groups may help to understand the pathogenic mechanisms in MND and may be useful for monitoring the effects of future trial treatment regimens. The research described in this thesis presents the development of a pattern recognition system based on neural networks to correctly distinguish between motor neurone disease (MND) patients and controls when presented with a nuclear magnetic resonance (NMR) spectrum. The NMR spectra are pre-processed to obtain consistent data, and statistical parameters are extracted and selected from each spectrum. Four statistical neural network classifiers are used to provide information and initial decisions (MND/normal). A neural network is then used to combine these to give a final decision. Experimental results indicate that the system can achieve high performance classification on the spectra, including spectra not seen by the system during training. The experiment was repeated on different training and test sets to validate the method and the repeated design shows that the final system was able to achieve high performance classification. A ifizzy rule-based system teclmique is applied to translate and extract rules encoded in weights of neural network classifiers. The neural networks are translated into a few comprehensible rules to understand how the network performs the final decision.