一种基于FCM和VBM的头部磁共振图像多组织分割算法

A method for the segmentation of multi-tissue from head MR images based on FCM and VBM

为了在无损情况下获得精确的人体头部三维模型,需要从颅脑的MRI序列图像中分割出头皮、颅骨、灰质、白质和脑脊液等5种主要的组织。首先,通过BET算法从原始的MRI图像中摘取出颅脑区域,然后通过模糊C-均值聚类(FCM)算法对得到的颅脑区域进行细分得到灰质、白质和脑脊液,继而再通过基于体数据的形态学(VBM)分割算法分割出颅骨、头皮和脑部空腔等非脑组织区域,最后对分割得到的各个组织进行平滑和形态学处理,最终成功分割出了所需的5种组织。通过与K均值聚类算法及形态学分割方法对比发现,该分割算法在组织边缘梯度更高的情况下具有更低的形态失真度。

In order to obtain the accurate three-dimensional model of human head through a non-invasion way,five main tissues named scalp,skull,gray matter,white matter and cerebrospinal fluid （CSF） should be segmented from MRI sequence images.Firstly,the brain tissue was extracted from the original MRI images by the brain extraition tool（BET） algorithm and then the brain region was segmented into gray matter,white matter and CSF by the fuzzy cluster method （FCM）.Secondly,the non-brain tissue was segmented into the scalp,skull and cavum by the voxel-based morphometry （VBM） segmentation algorithm.Thirdly,the obtained tissues were smoothed and morphologically processed to get the final five separate tissues.By comparing with K means cluster algorithm and the morphological segmentation algorithm it was found that the developed method gains the clearer edge and the lower shape distortion.