Abstarct: Multiple sclerosis (MS) is known as a chronic inflammatory disease of the Central Nervous System (CNS). In people with MS, immune system damage to the isolating laxyer of myelin around the nerve fibers in the brain, spinal cord and optic nerves that creates lesions and CNS atrophy in both Grey Matter (GM) and White Matter (WM) tissue. Magnetic Resonance (MR) imaging is one of the most important tools for diagnosis and monitoring of disease progression and treatment efficacy in patients with MS. But the detection and segmentation of MS lesion in MR images is challenging. Variability in lesion location, size, shape and anatomical variability between subjects are some factors that cause accurate identification of MS lesions in MR images extremely difficult. So many methods have been proposed to automatically segment MS lesions. The approaches have been classified between supervised and unsupervised methods. In this research, we try to use the advantages of the two approaches, using combination of mentioned methods. In this thesis, MS lesions were segmented by three classifiers Hidden Markov Random Field (HMRF), K-Nearest Neighbors (KNN) and Support Vector Machine (SVM) separately. The first one utilizes a statistical atlas to obtain an initial segmentation of the voxels belonging to different tissue classes. Then HMRF algorithm has been applied for segmenting of the brain into its different compartments on the T1 and the T2 sequences. From these segmentations, a threshold for the FLAIR sequence is automatically computed and postprocessing operations select the most plausible lesions in the obtained hyperintense. The second one focuses only on the segmentation of lesion and uses KNN and SVM classification. These methods use voxel location and signal intensity information for determining the probability being a lesion per voxel, thus generating probabilistic segmentation images on FLAIR sequence. By applying a threshold on the probabilistic images binary segmentations are derived. Finally by combining the results using a majority vote method, we have achieved a segmentation fault with the lowest. The performance of this algorithm is quantitatively evaluated on 20 MS patients that are provided by MS lesion segmentation grand challenge dataset (MICCAI 2008). The average value of Dice Coefficient Percentage (DSC) and Positive Predictive Value (PPV) are computed by spatially comparing the results of present procedure with expert manual segmentation. The values of DSC and PPV for purposed method are equal to 80.03% and 0.7661 respectively. The results showed acceptable performance for the proposed approach, compared to those of previous works.