Efficient fuzzy c-means based multilevel image segmentation for brain tumor detection in MR images

Image segmentation in MR images gives valuable information and plays a vital part in identifying the different kinds of tumor. Various learning techniques have been utilized for tumor detection by comparing extracted feature points of the image under study and reference image. However, it is a challenging task to build a reliable data for brain tumor detection by training due to large variations of brain image in terms of shape and intensity. This work focuses on edema and tumor segmentation that is based on skull stripping and kernel based fuzzy c-means approach. Clustering process is improved by combining multiple kernel based on the spatial information. Our multilevel segmentation approach relies on the global matching information between the image distributions and avoids the need for pixel wise information that reduces the computational complexity. Graphcut algorithm is incorporated in this framework as a co-segmentation to identify exact cut point between edema and tumor so that edema is removed from tumor. In this approach, clearer visualization of edema is possible and tumor is identified with extra space for proper removal. Simulation results reveal that our approach outperforms the other existing methods for complete tumor and edema segmentation.