2017 International Conference on Computer Science and Engineering (UBMK), Antalya, Turkey, 5 - 08 October 2017, pp.171-176
Skull extraction from Magnetic Resonance (MR) head image datasets is the process of segmentation of brain tissues from other tissues (e.g., skin, bone, fat) and has an important role in computer-assisted operations. Because, the accuracy of skull extraction affects the next stages of digital image processing or analysis. Semi-/full automated detection and extraction of the skull from MR datasets is a difficult task due to inhomogeneous intensity values, overlapping of non brain and brain tissues, noise, unclear edges, existence of lesions or tumors. Also, skull size, shape and thickness may vary in each person. Therefore, gray level based or shape based skull detection and extraction algorithms can not always yield in desired results. To overcome these difficulties, in this paper, a hybrid algorithm is proposed. In this approach, anatomy knowledge, binary morphological operations and Gaussian model based classification is used. Experimental results show efficiency of the proposed approach.