Genetic Algorithms for Finite Mixture Model Based Tissue Classification in Brain Mri

Finite mixture models (FMMs) are an indispensable tool for unsupervised classification in brain imaging. Fitting a FMM to the data leads to a complex optimization problem. This optimization problem is difficult to solve with standard local optimization methods (e.g. by the expectation maximization (EM) algorithm) if a good initialization is not available. In this paper, we propose a new global optimization algorithm for the FMM parameter estimation that is based on the real coded genetic algorithms. Our specific contributions are two-fold: 1) We propose to use blended crossover in order to reduce the premature convergence problem to its minimum. 2) We introduce a completely new permutation operator specifically meant for the FMMparameter estimation. We demonstrate the good behavior of our algorithm compared to the EM-algorithm and a standard real coded genetic algorithm with the tissue classification task within the magnetic resonance brain imaging. Phantom images as well as real three dimensional image data with pathology are considered. The tissue classification results by our method are shown to be consistently more reliable and accurate than with the competing parameter estimation methods.