中文摘要功能性磁共振(fMRI)已成为脑科学研究的重要手段和工具。它具有无侵入、
无伤、高速、高分辨率、可同时获得结构与功能图像等一系列优秀性能,被广泛
应用于脑科学实验及临床研究。
本文主要采用国外采集的一些 fMRI 数据,首先使用SPM软件进行预处理并分
析脑激活区,随后试图利用模式识别的一些经典方法找到大脑体元与图像刺激物
之间的潜在关系。由于人类全脑区的数据量十分巨大,需要对于已经检测出的激
活区,结合生理学先验知识进行特征区域选择。选取特征区域后,分别建立了基
于支撑向量机(support Vector machine,SVM)和核岭回归(Kernel Ridge
Regression)的模型,使用一部分已知的数据进行训练,然后再利用该模型对其
余数据进行预测,从而实现脑部活动解码。最后分析其性能、优化方法提高准确
率。实验结果表明,这种解码方法具有一定的可靠性和稳定性。6482
关键字:功能核磁共振成像,SPM,支撑向量机,核回归,脑活动解码 毕业设计说明书(论文)外文摘要
Title Research on decoding of brain activity state based on fMRI
Abstract
Functional magnetic resonance imaging (fMRI) has become an important
method and tool in the field of brain science. It is non-invasive, harmless,
with high speed and high resolution. It can also acquire structural and
functional image at the same time. So it is widely used in the experience
of brain science and clinical research.This article adopts the fMRI data
collected abroad. First, I use the SPM software to preprocess and analyze
brain activation regions. Then I attempted to utilize some classic methods
in pattern recognition to find the potential relationship between brain
voxels and image stimulus. Because the data volume of the whole brain is
quite huge, so we need to do feature extraction by utilizing physiology
prior knowledge to the detected activation regions. After choosing the
feature ,I build the model based on support Vector machine(SVM) and Kernel
Ridge Regression. I use some of the acquired date to train the classifier,
then I use the classifier to do the prediction of the other date. That is
how I realize the decoding of brain activity. After that, I analyze the
performance and optimize the method to improve the accuracy. The result
shows that this decoding method is reliable and stable.
Keywords fMRI, SPM, support Vector machine(SVM), Kernel Regression,
Brain activity decoding 本科毕业设计说明书(论文) 第 I 页,共Ⅱ页
目 次
1 绪论 1
1.1 人脑基本生理结构 ... 1
1.2 功能核磁共振技术 ... 1
1.2.1 BOLD 信号对比度机制 ... 1
1.2.2 血液动力响应学生理机制 2
1.2.3 功能核磁共振成像特点.. 3
1.3 本课题国内外研究现状 ... 3
1.4 本文研究内容和组织结构 . 5
2 实验设计及数据预处理 6
2.1 模块设计 ... 6
2.2 事件相关设计 ... 7
2.2.1 多模式人脸的实验设计.. 7
2.2.2 有关食物、金钱、小杂货的三类物品的实验设计 7
2.3 SPM 数据预处理.. 8