QQ客服 官方微博 反馈留言

题名	计算神经解剖学中的分割算法研究
作者	朱朝喆
学位类别	工学博士
答辩日期	2004-06-01
授予单位	中国科学院研究生院
授予地点	中国科学院自动化研究所
导师	蒋田仔
关键词	解剖学 分割
其他题名	MR Image Segmentation in Computational Neuro-anatomy
学位专业	模式识别与智能系统
中文摘要	大脑的结构是其复杂功能的物质基础，脑组织与结构的分割是计算神经解 剖方法学研究的核心内容之一。由于MRI图像中的各种干扰与大脑自身复杂性 等因素的存在，分割已成为一个典型的“病态问题”，是计算神经解剖学甚至整 个脑科学研究与应用道路上的瓶颈。作者针对计算神经解剖学中“脑组织分割、 神经结构分割和脑病变区分割”三个基本问题进行了深入细致的研究，提出了 相应的计算理论与算法，并将这些算法应用到神经科学的实际问题研究当中。 论文的主要工作可以简单总结如下： （1）．对脑组织分割问题进行深入分析，建立了脑组织分割问题的系统描述。 （2）．提出脑组织属性分布的空间不平稳性概念与局部脑组织分布模型，丰富 了MRI脑组织分割理论。 （3）．提出了多上下文模糊聚类（MCFC）算法，解决了MRI脑组织分割研究 中灰度不均匀性这一经典难题。 （4）．提出了基于解剖知识的多上下文模糊聚类（adMCFC）算法，进一步完 善了原有MCFC的性能。 （5）．提出了基于灰度直方图统计的FCM快速实现算法，显著加速了MCFC 脑组织分割的速度。 （6）．基于MCFC提出了多通道信息融合MS病变分割算法。 （7）．提出了两种基于扩散张量成像（DT-MRI）的白质束分割算法 作者还与北医三院合作将MCFC脑组织分割算法应用到强迫症患脑组织定 量分析研究当中，得到了强迫症患者的脑白质归一化体积显著大于正常对照的 结论，为强迫症患的临床诊断提供参考指标。
英文摘要	Structure is the basis of function. Segmentation, which plays an important role in computational neuro-anatomy, is an ill-posed problem caused by both arifacts in MR images and complexities of the brain itself. Concepts, theories and algorithms are proposed in the dissertation to describe, model and solve the problems in segmentation of brain tissues, lesions and sub-cortical structures in MR images. Main contributions of the dissertation can be summarized as follows. (1). A complete mathematical description of the problem of MR/brain tissue segmentation was set up on the basis of neuroanatomic coordinations and four spaces of labeling transformation, biophysical properties, theoretical MRI signal and observation MILl signal. (2). A concept of instability of spatial distributions of brain tissues properties and a local distribution model of brain tissues were presented to extend the theory of MRI tissue segmentation. (3). Multi-context fuzzy custering (MCFC) was proposed and successfully solve the issue of intensity inhomogeneities, an open proplem in MRI brain tissue segmentation. (4). Anatomy-dependent MCFC (adMCFC) was proposed to improve original MCFC by adjusting the context size according to context's location in the brain. (5). A histogram analysis based fast implementation of fuzzy C-means (FCM) had been proposed to accelerate MCFC significently. (6).Two methods were proposed to isolate anatomically distinct white matter fascicules from diffussion tensor MRI(DT-MRI)data:Propagable region-growing(PRG)and Deformable model based segemntaion. (7).A knowledge-guided fusion framework was proposed on the basis of MCFC algorithm and information fusion Structure is the basis of function. Segmentation, which plays an important role in computational neuro-anatomy, is an ill-posed problem caused by both arifacts in MR images and complexities of the brain itself. Concepts, theories and algorithms are proposed in the dissertation to describe, model and solve the problems in segmentation of brain tissues, lesions and sub-cortical structures in MR images. Main contributions of the dissertation can be summarized as follows. (1). A complete mathematical description of the problem of MR/brain tissue segmentation was set up on the basis of neuroanatomic coordinations and four spaces of labeling transformation, biophysical properties, theoretical MRI signal and observation MRI signal. (2). A concept of instability of spatial distributions of brain tissues properties and a local distribution model of brain tissues were presented to extend the theory of MRI tissue segmentation. (3). Multi-context fuzzy custering (MCFC) was proposed and successfully solve the issue of intensity inhomogeneities, an open proplem in MRI brai
语种	中文
其他标识符	813
内容类型	学位论文
源URL	[http://ir.ia.ac.cn/handle/173211/5824]
专题	毕业生_博士学位论文
推荐引用方式 GB/T 7714	朱朝喆. 计算神经解剖学中的分割算法研究[D]. 中国科学院自动化研究所. 中国科学院研究生院. 2004.

个性服务

查看访问统计

相关权益政策

暂无数据

收藏/分享

除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。