Brain tumor is an abnormal mass of tissue with uncoordinated growth inside the skull, which may invade and damage nerves and other healthy tissues. Magnetic resonance imaging (MRI) device has been widely used in biological application, diagnostics, and detecting tumors. In this thesis, image segmentation is applied on MR images to detect brain tumors. These images were real human brain images taken from ALKAHDEMIA HOSPITAL. The segmentation process was done using two methods, fuzzy c-mean(FCM) clustering method and neural network based on self-organization map method (SOM). Brain images (real cases) of gray levels were divided into four clusters by segmentation; the clusters formation are based on a certain features that are related to the intensity of the cluster pixels in the whole image. With implementing the above mentioned methods on the brain images, the tumor location, were identified in each case. Comparison is made between these methods based upon efficiency of clustering, execution time and output information of the segmented images. The results of segmentation were in a very good agreement with the diagnosis of ALKADIMIA HOSPITAL doctor's reports. Computer programs have been designed, written and implemented in this work using MATLAB 7, for the previously mentioned methods of segmentation. This made the results very close to the real location and values of the referenced tumors.