脑图像数据中的独立分量分析方法  

Independent Component Analysis Method in Brain Image Data

作　　者：马斌[1] 陈俊杰[2] 

机构地区：[1]山西中医学院医药管理学院,太原030024 [2]太原理工大学计算机与软件学院,太原030024

出　　处：《计算机工程》2014年第3期205-207,共3页Computer Engineering

基　　金：国家自然科学基金资助项目"基于fMRI的个性化图像情感标注及其本体库研究"(60970059)

摘　　要：针对脑功能磁共振成像在处理数据时空间维数较大的问题,提出一种空间独立分量分析(ICA)方法。研究空间ICA方法的基本模型结构和空间ICA的3种常见算法,即Infomax算法、Fixed-Point算法和Orth-Infomax算法。设计中文词义辨别实验,并使用线性相关方法进行算法比较。实验结果表明,与Infomax算法、Fixed-Point算法相比,Orth-Infomax算法任务相关分量的时间序列与参考函数的平均相关系数最大,具有较高的求解质量和求解效率,能够有效处理脑功能磁共振成像系统中存在的大量数据。Independent Component Analysis（ICA） is an effective method of data processing of the brain functional Magnetic Resonance Imaging（fMRI）. Aiming at the feature that the spatial dimension of fMRI data is large, spatial ICA is selected to be discussed. The basic model structure of ICA and the three common algorithms of spatial ICA are deeply researched, including Infomax algorithm, Fixed-Point algorithm and Orth-Infomax algorithm. Chinese word meaning differentiation experiment is designed and analyzed with the linear correlation method. Experimental results show that, the time series of CTR in the Orth-Infumax algorithm has the maximum average correlation coefficient with the reference function, compared with Infomax algorithm and Fixed-Point algorithm, which has the high quality of the solution and the solving efficiency and can efficiently process the fMRI system data.

关 键 词：脑功能磁共振成像 独立分量分析 一致任务相关成分 正交信息极大化算法 源信号 线性相关 

分 类 号：TP393.41[自动化与计算机技术—计算机应用技术]