Fuzzy Neural Networks and Genetic Algorithms for Medical Images Interpretation

In this paper, we propose an approach for detection and specification of anomalies present in medical images. The idea is to combine three metaphors: neural networks, fuzzy logic and genetic algorithms in a hybrid system. The neural networks and fuzzy logic metaphors are coupled in one system called fuzzy neural networks. The genetic algorithm adds to this hybridizing the property of total research like an initialization of the fuzzy neural networks training algorithm witch is based on an adapted version of the back propagation algorithm. After applying the growing region algorithm to extract regions, the fuzzy neural network detect the suspect regions, which are interpreted by the fuzzy neural network of specification. Some of experimental results on brain images show the feasibility of the proposed approach