一种基于马尔可夫随机场的弥散张量成像(DTI)图像分割的新算法  

A Novel DTI Image Segmentation Algorithm Based on Markov Random Field

作　　者：彭洁[1] 徐启飞[2] 王华峰[3] 吕庆文[1] 冯衍秋[1] 陈武凡[1] 

机构地区：[1]南方医科大学生物医学工程学院,广州510515 [2]临沂市人民医院放射科,山东临沂276003 [3]南方医科大学基础医学院,广州510515

出　　处：《医疗卫生装备》2012年第4期12-13,共2页Chinese Medical Equipment Journal

基　　金：国家重点基础研究发展规划(973计划)项目(2003CB716102);国家自然科学基金重点项目(30730036)

摘　　要：目的:提出一种基于马尔可夫随机场(MRF)的弥散张量成像(DTI)图像分割的算法。方法:利用马尔可夫随机场模型,挖掘图像中的弥散张量信息,根据贝叶斯定理将图像分割问题转化为最小后验能量的求取,运用迭代条件模型求解。结果:该算法对DTI图像分割效果明显优于K均值算法,且该效果亦优于该算法对常规MRI T_2WI图像的分割效果。结论:该算法能够充分利用弥散张量矩阵蕴含的空间上下文信息实现DTI图像的有效分割。Objective To propose a novel Markov random field（MRF） based segmentation algorithm for diffusion tensor images（DTI）.Methods The distance measure defined by Frobenius norm was introduced in order to utilize more spacial information of the diffusion tensor matrix of image voxels.The segmentation issue was transformed to the Minimum A Posteriori（MAP） by Beyes theorem,and the Iterative Conditional Model（ICM） algorithm was employed to achieve the solution of latter MAP problem.Results The comparison of segmentation results between the proposed algorithm and Kmeans segmentation algorithm for DT-MRI image was made,which indicated that the proposed algorithm could segment the DTI images more accurately than the K-means algorithm.Moreover,with the same segmentation algorithm of MRF,better outcomes were achieved in DTI image than that in conventional MRI T2WI image.Conclusion The proposed algorithm can adequately utilize spacial information contained in voxel＇s diffusion tensor matrix to achieve the efficient segmentation of DTI images.

关 键 词：弥散张量 磁共振成像 图像分割 马尔科夫随机场 

分 类 号：R445.2[医药卫生—影像医学与核医学]