Evolutionary algorithms and neural networks applied to the computer - aided medical diagnosis

The purpose of presented work is to create a project and computer implementation of complex decision support system used in an important medical field, which is cardiology. This system is applied to support physical diagnosis concern different kinds of myocardial infraction. The system - called NEUROGEN v.01, is a kind of hybrid system, which is a combination of Genetic Algorithm (GA) and Neural Network (NN). The idea of this specific combination is that GA is used as a evolutionary method of learning of NN. In accordance with this special task, the NN is a three-layer feedforward network with eight numbers of input neurons, six numbers of hidden and five number of output neurons. The number of neurons in each layer was appointed on the base of data of the task. In this work, the purpose was to look for the optimal values of the parameters of algorithm, which are: crossover probability, mutation probability, the number of individuals in population, the number of generations of the algorithm and λ - parameter of function of activation which characterize neurons in NN. An extra task is to check if the beginning population has any influence on effectiveness of the system. In this paper there will be presented the way of rising of NEUROGEN v.01 and achieved results.