MR cranial image segmentation - a morphological and clustering approach

Computer-aided visualization of the brain surface has numerous applications in structural and functional brain mapping for neuroscience, as well as in surgical path planning. Surface visualization of the brain using magnetic resonance (MR) images requires pixels in the images to be discriminated, or segmented into various tissue types. In this work, an automated and efficient method, based on mathematical morphology and unsupervised clustering analysis, is proposed for segmenting 3-D MR images of the human head. The proposed method provides better rendered views of the brain surface with much less computation time, compared to a previous, direct fuzzy-based approach. The resulting rendered views of the brain surface are shown, as well as the efficiency and error analysis of the proposed methodology.