An algorithm of automatic human brain identification in complex CT images

Abstract

A method of automatic detection of human brain region in computed tomography (CT) images is described in this article. The main attention is paid to complex CT slices obtained in temporal or nasal level. Method is suitable for identification of region of interest (ROI) after surgery operations of brain as well. In cases mentioned above brain region may be fragmented; it can have contacts with other tissues with similar X-ray absorption properties; it may be not covered by bones as well. In such cases automatic brain region identification requires more sophisticated algorithms. The algorithm described below uses filtering and morphological approaches for identifying the biggest grey particles and particle growing technique for selection of correct areas. Experimental results were obtained that show the algorithm in the most cases completely removes tissues of skin, eyes and bones. The accuracy of brain identification was evaluated by comparison with areas defined by experts. The method was suggested to use as the first stage of image processing in automatic brain disease detection software.