A Method for Optimization of Fuzzy Reasoning using Genetic Algorithms and Its Application to Recognition of Medical Images

Abstract

We describe a method for optimizing the membership functions (MSF's) of fuzzy rules using genetic algorithms (GA's) and apply this GA - based fuzzy reasoning to discriminate heard diseases from echocardiograms. The method is to use Gaussian - distributed MSF's obtained from texture features of the echocardiograms. The standard deviations of the MSF's acting as parameters are optimized through training process using the GA. In the GA - based training, a two-step fitness function for selection phase is employed in order to increase the accuracy of the classification. The results of our experiments are very promising.