【作者】 罗云；

【导师】 王秀坤；

【作者基本信息】 大连理工大学 ， 计算机软件与理论， 2006， 硕士

【摘要】 了解脑的功能是21世纪科学的重大挑战之一。目前的“人类脑计划”旨在加强脑功能的基础研究,并开发用于分析、整合、合成、建模、模拟与提供各种数据的工具。越来越突出的青少年心理问题正在被世界各国所关注。而其中主要以焦虑、抑郁等神经症行为的增多为主。本文研究青少年焦虑抑郁在人脑中的表现,包括正常时和受刺激时的表现,以揭示人脑的高级功能。 本文论述了青少年焦虑抑郁SPECT数据处理的一般方法。在脑功能成像研究过程中,除精确的设计实验、选择被试和采集数据外,最关键的是处理、分析和统计。目前脑功能成像研究大多着重于脑区功能定位,即单纯确定哪些脑区参与了认知加工,甚至精确到某一Brodmann分区或皮下结构中的不同位置,其方法是对比控制时的信号与任务时的信号,判断两者是否有显著性的差异。 本文将支持向量机和关联规则用于青少年焦虑抑郁SPECT数据挖掘。建立一个可以在被测试者之间分辨焦虑抑郁状态的分类器。我们沿着两种不同的思路对脑图像进行特征提取:一种是沿用基于内容的图像检索方法,使用了纹理、形状、色彩,以及它们的联合特征;另一种是使用基于感兴趣区和t值最强点的特征提取。最后用支持向量机的理论对SPECT数据进行分类。将大脑用树状结构表示为各个脑区,然后使用关联规则挖掘出不同脑区之间的关系。 本文对基于知识库的脑功能成像诊断模型进行了初步探索。根据脑功能成像的特点,引入知识工程的知识表示理论,通过一种基于知识的树状表示法描述该问题,并在此基础上表示辅助诊断中的规则,从而形成问题的知识库。提出一种基于Prolog语言表示的知识库建立的方法,从而实现脑功能成像辅助诊断模型。更多还原

【Abstract】 To understand the brain and its function is one of the great challenges in 21 century. Now human brain project aims to research the brain function and develops all kinds of tools to analyze and compose the brain and etc. There are more and more mental problems of young people which are paid attention to by the whole world. Number of anxiety disorder and despondent patients increase rapidly. This paper researchs the brain data of despondent patients including normal state and provocative state, in order to find out the high level function and secret of brain.This paper is involved with general measures of SPECT’s brain data. In the research of brain image, it is important to process, comform, compose, model, simulate, analyze except for accurately designing experiments, picking up testees and collecting data. Now it mostly focuses on the relations between brain areas and brain functions and simply make sure which part of brain can be functional. The method is to compare the signal of control and mission.This paper uses support vector machines and association rules to min anxiety disorder’s SPECT data. It builds a classifier to distinguish testees who are anxiety disorders or not. There are two different ways to extract the brain image features, one uses image semantic classification with texture, shape and color. The other uses region of interest and the max value of t. Then it uses SVMs to classify the SPECT data. After representing brain with a tree-like structure, it uses association rules to min the relations between different brain areas.Finally this paper talks about a knowledge-based brain imaging diagnosing model. According to the properties of brain image, a knowledge-based tree-like representation is introduced to present brain imaging diagnosing model. And the knowledge that is formed by the rules is used with the database and feature extraction to realize the diagnosing model.更多还原