Hidden Markov Random Fields and Particle Swarm Combination for Brain Image Segmentation

The interpretation of brain images is a crucial task in the practitioners' diagnosis process. Segmentation is one of key operations to provide a decision support to physicians. There are several methods to perform segmentation. We use Hidden Markov Random Fields (HMRF) for modelling the segmentation problem. This elegant model leads to an optimization problem. Particles Swarm Optimization (PSO) method is used to achieve brain magnetic resonance image segmentation. Setting the parameters of the HMRF-PSO method is a task in itself. We conduct a study for the choice of parameters that give a good segmentation. The segmentation quality is evaluated on ground-truth images, using the Dice coefficient also called Kappa index. The results show a superiority of the HMRF-PSO method, compared to methods such as Classical Markov Random Fields (MRF) and MRF using variants of Ant Colony Optimization (ACO).