An evolutionary HMRF approach to brain MR image segmentation using clonal selection algorithm

The Hidden Markov random field (HMRF) model provides the basis for a spatially constrained clustering scheme, and hence has been widely used for image segmentation. In many HMRF-based segmentation approaches, the statistical parameters involved in this model are estimated by using the expectation maximization (EM) algorithm, which, however, is commonly acknowledged to have several drawbacks, such as the over-fitting and local convergence. Recently, the clonal selection algorithm (CSA), a novel immune-inspired evolutionary optimization tool, has been increasingly used to replace local search heuristics in various applications, due to its proven global, parallel and distributed search ability. In this paper, we incorporate the CSA into the HMRF model estimation, and thus propose the evolutionary HMRF (eHMRF) algorithm to delineate the gray matter, white matter and cerebrospinal fluid in brain magnetic resonance (MR) images. This segmentation algorithm has been compared to the state-of-the-art GA-EM algorithm and the HMRF-EM segmentation function in the FMRIB Software Library (FSL, Version 2008) package in simulated brain MR images. Our results show that the proposed eHMRF algorithm can differentiate brain structure more effectively and produce more accurate segmentation of brain MR images.