摘要
扩散张量磁共振成像(DT-MRI,简称DTI)是一种能够对脑组织中复杂的白质纤维束进行观察和追踪的一种特殊形式。而针对DTI对噪声和灰度敏感,传统的分割方法难以对脑组织进行精确分割的特点,设计了一种基于图像相似度的多权重图谱DTI自动分割算法(MAPS)。MAPS是在MAIS的基础上通过分割图像的自相似性及引导滤波进行优化的方法。结果为MAPS分割的重叠率要优于MAIS大约6%,并且MAPS的分割结果比MAIS在三维效果上更接近专家的手动分割结果(即金标准),结果表明MAPS相较于MAIS有着更精确的分割效果及更优异的性能。
Diffusion tensor magnetic resonance image(DT-MRI,referred to as DTI)is one main kind of method to observe and track complex white matter fiber bundles in brain tissue.In view of the fact that DTI is easily susceptible to the inhomogeneous intensity and noise,traditional segmentation methods are difficult to segment brain tissues accurately.Multi-weight-atlas based DTI segmentation with image similarity(MAPS)is proposed,MAPS is based on the MA-IS method,in which the weight obtained by MA-IS algorithm is optimized by the self-similarity of the image to be segmented,and the boundary of segmentation result is optimized with the guided filtering technique.Experimental results are that the overlap ratios of the MAPS segmentation are improved about 6%compared with that of MAIS,and the segmentation results of MAPS algorithm are closer to the manual segmentation results than MAIS in a stereo-vision effect.The results show that MAPS has more accurate segmentation effect and better performance than MAIS.
作者
李国琴
王瑾
谭艳丽
焦冬莉
薄晓宁
刘继军
Li Guoqin;Wang Jin;Tan Yanli;Jiao Dongli;Bo Xiaoning;Liu Jijun(Taiyuan Institute of Technology,Taiyuan 030008,China)
出处
《电子测量技术》
2020年第6期116-122,共7页
Electronic Measurement Technology
基金
基于多权重图谱的DTI自动分割算法研究(2018LG08)项目资助。
关键词
扩散张量磁共振图像
相似度
多图谱
多权重
自动分割
diffusion tensor magnetic resonance image
similarity
multi-atlas
multi-weight-atlas
automatic segmentation
