We propose a framework for automated segmentation of Multiple Sclerosis (MS) lesions from MR brain images. It integrates a priori tissues and MS lesions information into a Graph-Cuts algorithm for improved segmentation results. We formulate the energy terms to include a priori information as well as the information derived directly from the MR images. We validate our method on a dataset of 37 MS subjects with a broad range of lesion loads. Results indicate that integrating a priori information with the information derived from the images can improve the segmentation outcome.