Image segmentation is the process of partitioning an image into smaller non-overlapped and meaningful regions based in part on some homogeneity characteristics. Many high-level processing tasks such as feature extraction, object recognition and medical diagnosis depend heavily on the quality of solutions. In medical image analysis, images usually contain some artifacts such as noise, image volume effect and bias field effect due to various factors, for instance, environment and acquisition devices, and have complex structures. Therefore, image segmentation remains a difficult task even if various techniques and methods of different accuracy and degree of complexity have been introduced in the literature. Several approaches such as fuzzy clustering, region-based active contour, Markov random field, have been found that can produce promising results; however, still many key open issues remain to be investigated. Up to now, there is no gold standard method and segmentation procedures still need a significant amount of expert intervention for improving the performance.Metaheuristics are a high-level procedure designed to solve optimization problems by the process of searching optimal solutions to a particular problem of interest. Metaheuristics are generally applied to problems for which there is no satisfactory algorithm able to solve them effectively. Therefore, they are widely used to solve complex problems and have proven to be successful in many fields of application with varying degrees of success. Considering the image segmentation problem as one of the optimization problems solved by metaheuristics, image segmentation has attracted many researchers in recent years. In many successful applications, it can be seen that the traditional approaches for image segmentation have been combined with metaheuristics in different perspectives in order to improve their performance.Bearing those in mind, we propose in this work three image segmentation methods for magnetic resonance (MR) brain images based on mono-objective, multi-objective and hybrid metaheuristic optimization techniques. In each method, first, the basic model for the image segmentation problem is extended to incorporate more image information (spatial or spectral) such that more and better characteristics in segmented image can be achieved. Then, metaheuristic algorithms are adapted or developed to take place in optimization step. The proposed methods were evaluated on both simulated MR images and real MR images and compared with a set of recent methods in the literature. The obtained results show clearly the efficiency of the proposed ideas.Keywords: Image segmentation, fuzzy clustering, region-based active contour, Markov random field, metaheuristics, multi-objective optimization, hybrid metaheuristic, MRI.