Neurological diagnosis system based on neural networks

11 ABSTRACT Master of Science Thesis NEUROLOGICAL DIAGNOSIS SYSTEM BASED ON NEURAL NETWORKS MUZAFFER DO?AN Anadolu University Graduate School of Natural and Applied Sciences Computer Engineering Program Supervisor: Assoc.Prof.Ahmet BABANLI 2003, 99 pages The main issue in preparation of an expert system is to get the expert person's knowledge and to put this knowledge into an available form to use for computer programs. In this thesis, a study on listing and classification of headache symptoms was done in order to make it easy to write computer programs which diagnose headache diseases. After this classification, classification of headache disease using Neural Networks was tried. Some of the Neural Networks models which are perceptron networks, self organizing mapping and backpropagation network models are applied to a narrow group of headache diseases and the results were compared. In the diagnosis problem, the mostly encountered diseases which are migraine with aura, migraine without aura and tension type headache were examined. As a result, it was seen that the most efficient neural network method is perceptron networks in diagnosis of headache diseases. Future reference for headache classification works was also prepared. Keywords: Neural Networks, Headache Diseases, Expert Systems, Diagnostic Problems.