Segmentation of Brain Regions by Integrating Meta Heuristic Multilevel Threshold with Markov Random Field

Image segmentation is a method of delineating prominent structures in an input image for post processing such as texture based analysis, feature extraction, selection, and classification. Various medical applications rely on efficient image segmentation algorithms for disease diagnosis and management. In this work, autonomous segmentation of brain regions of Magnetic Resonance Images (MRI) is attempted for neuroimaging applications. It is carried out by combining heuristic based image segmentation with stochastic modeling. Initially, Particle Swarm Optimization (PSO) algorithm is used to identify ideal intensity thresholds. Labeling priors from PSO are derived for the Markov random field. Markov Random Field (MRF), a probabilistic approach is used to improve the partitioning of the initial segmented image obtained by the multilevel threshold technique. The efficacy of the segmentation procedure is improved by refining the obtained apriori information by Maximum a posteriori (MAP) estimation of the MRFMAP model. By this approach, the spatial information is incorporated into the segmentation process using MRF. A new metaheuristic MRF image segmentation technique is proposed here to take advantage of both the methods. Performance assessment of the proposed method is carried out using a numerical metric that evaluates the silhouette index of the estimated clusters. Experiments conducted on different MRI datasets show the proposed methodology produces an average improvement in cluster classification of 4.42% in terms of silhouette index for clinical datasets.