Automatic Head MRI segmentation combining FCM and VBM

In the diagnosis and operation of the brain, there are five kinds of main tissues, such as gray matter, white matter, cerebrospinal fluid, scalp and skull, those need to be segmented from the MRI to sequel accurate three-dimensional head model. Aiming at this problem, this paper proposes a segmentation method combining FCM & VBM algorithm. After the craniocerebral regions were extracted from the original MRI images by the BET algorithm, the brain regions were subdivided to obtain gray matter, white matter and cerebrospinal fluid by FCM algorithm. Then, the skull, scalp and brain were segmented by VBM segmentation algorithm. And then the smooth and morphological treatment of the separated tissues was carried out. Finally, five kinds of tissues were obtained. Compared with K-means clustering algorithm and morphological segmentation method, it is found that the segmentation algorithm has lower morphological distortion in the case of higher edge gradient. Introduction The brain as the most important and sophisticated nerve center of the brain, the brain of a variety of disease detection and brain surgery require high precision, so the head of the various operations before the need for precise pre-analysis and lesion location. For the pre-operative detection, it is very important to get the 3D model of the head, which is very important for the diagnosis and operation navigation of brain diseases. At present for the head area imaging technology mainly CT and magnetic resonance imaging (MRI), which because of MRI for soft tissue has a higher resolution often become the first choice for head lesions. However, because of the complexity of the brain tissue and the similarity of gray levels in the MRI images such as cerebrospinal fluid, skull and head cavities, there are still considerable difficulties in segmenting the various tissues from MRI images. From the literature that has been reported in recent years, there are a variety of algorithms for brain MR image segmentation[1; 2].But these algorithms are mostly for specific areas, such as the extraction of the brain or from the extracted brain region is divided into gray matter, white matter or tumor, etc. [3; 4], and for the head MRI image of the overall segmentation , Especially one-time from the original head MRI image segmentation contains the scalp, skull, cerebrospinal fluid, gray matter, white matter and even the head cavity, including a variety of organizations, there is no universal and effective method.In the segmentation algorithm proposed in recent years, the more prominent head MRI segmentation methods are more classic: Dogdas et al [5] using pure mathematical morphology method from the head MRI image segmentation skull model. The method has the advantages of a simple and practical process, but the disadvantage is due to the skull on the division of the larger scale smoothing led to a large skull shape distortion, in addition to the method is difficult to deal with mandibular parts of the skull. Huang et al [6] proposed a voxel-based morphometry (VBM) method for segmentation of cranial magnetic resonance images. Although this method can segment brain tissue and skull, Cavity and other parts. its shortcomings are divided after the existence of hollow cerebrospinal fluid, there are discontinuous points in the skull, the brain is not identified voxel. 2nd International Conference on Automation, Mechanical Control and Computational Engineering (AMCCE 2017) Copyright © 2017, the Authors. Published by Atlantis Press. This is an open access article under the CC BY-NC license (http://creativecommons.org/licenses/by-nc/4.0/). Advances in Engineering Research, volume 118