一种边界和马尔可夫随机场相结合的脑MRI医学图像分割方法  

Brain magnetic resonance image segmentation combined boundary and Markov random field

作　　者：林江[1] 戴齐[1] 欧阳婷雪 鞠斌[2] 邹翎[2] 

机构地区：[1]西南交通大学信息科学与技术学院,四川成都611756 [2]四川大学华西医院,四川成都610041

出　　处：《中国医学物理学杂志》2015年第5期717-720,共4页Chinese Journal of Medical Physics

基　　金：成都市科技支撑项目(11PPYB109SF;2014-HM01-00314-SF)

摘　　要：目的:在现有的脑MRI医学图像分割方法基础上,以阿尔茨海默病(Alzheimer's Disease,AD)数据为例,提出了一种基于边界和马尔可夫随机场为基础,再结合李代数和流场论对图像进行分割和配准的方法。方法:该方法以基于边界的分割方法去除颅骨和非组织,再利用马尔可夫随机场分割脑组织,最后结合李代数和流场论对图像进行标准配准,并与当前最为常用的SPM-VBM方法进行比较。结果:以现有的AD病人脑MRI数据为基础进行分析对比,能够得到更加有效的脑组织分割和更精确的脑激活区定位。结论:该方法能够显著提高分割效果。Objective Based on the existing medical image segmentation methods for brain magnetic resonance image （MRI）, taking the Alzheimer＇s disease （AD） data as the example, a method based on boundary and Markov random field （MRF）, combined with Lie algebra and flow field theory, is proposed for image segmentation and registration. Methods The segmentation based on boundary was firstly used to remove skull and non-organization, and then, a MRF was used to segment the brain tissue, and finally combining Lie algebras with flow field theory, the standard registration was carried out for the images. The results were compared with the results of the mostcommon statistical parametric mapping-voxel based morphometry. Results The comparative analysis of the AD patients＇ brain MRI data was more effective in segmenting brain tissue and locating brain activation areas. Conclusion The proposed method can significantly improve the segmentation effect.

关 键 词：MRI 医学图像分割 边界 马尔可夫随机场 阿尔茨海默病 

分 类 号：R312[医药卫生—基础医学]