Magnetic Resonance Based Ventricle System Classification by Multi-Species Genetic Algorithm

Early detection of medical abnormalities is an essential component in the accurate diagnosis and treatment of disease and disability. Automatically isolating the structure of the brain's ventricle system has the potential to assist neurologists in identifying abnormalities such as tumors, hydrocephalus and damage due to stroke. In this paper we present a multi-species genetic algorithm for isolating the ventricle system from multiple layers of brain images produced by magnetic resonance imaging (MRI) technology. Significant research has been performed on automatic classification of brain components from MRI images. To the best of our knowledge, this is the first approach to utilize multi-species genetic algorithms as a means of classifying brain components from MRI scans whereby the genomes control vertical and horizontal edge detection and locality based information suppression. This method requires a simple set of input data to be used for training the genetic algorithm. The results illustrate an effective method for isolating a patient's ventricle system