【作者】 张伟伟；

【导师】 唐焕文；

【作者基本信息】 大连理工大学 ， 运筹学与控制论， 2006， 硕士

【摘要】 用功能磁共振成像(functional magnetic resonance imaging,fMRI)技术来研究脑的功能,仪器记录的只是脑皮层不同位置的活动,这些活动是脑的深层活动所产生的各种信号的迭加,这些独立的源信号反映了脑的实际的活动情况。把这些独立的源信号分离出来可以获得真正意义的生理信号,从而有助于了解大脑的活动情况,进而为解决实际问题服务。而独立成分分析(independent component analysis,ICA)就是解决这类问题的一种有效的方法。本文围绕利用独立成分分析处理脑功能磁共振成像数据做了以下工作: 1.采用一种新的ICA算法(new fixed-point,NewFP)来处理fMRI数据,并且利用信息论的方法AIC(Akaike’s information criterion)或MDL(minimum description length)估计源信号的个数,用主成分分析(principal component analysis,PCA)对数据进行降维,通过这样的处理,有效地估计了源信号的个数,滤去了大量的冗余信号,大大提高了计算速度。将NewFP算法及FastICA算法用于处理实际的fMRI信号,比较发现,NewFP算法在估计激活的时间动力学准确性上优于FastICA算法。 2.由于在实际实验当中,通常需要的是数据处理的统计的结果,文章中介绍并分析了目前国际上比较流行的三种用ICA来处理多个被试的fMRI数据的方法——groupindependent component analysis(Group ICA),这对于处理大型的功能磁共振成像数据非常必要。 3.利用NewFP算法结合比较有效的一种Group ICA的方法对成组的fMRI数据进行处理,统计结果进一步表明NewFP算法的准确性,并且结果表明利用这种Group ICA的方法处理多个被试的fMRI数据不仅可以减少计算量,提高计算速度,而且可以保证结果的准确性。更多还原

【Abstract】 When people study the function of the brain using the functional magnetic resonance imaging (fMRI), instrument only notes the activation of the different areas of the cortex which are the superposition of the different signals produced from the deep activation of the brain. These independent signals reflect the actual situation of the activation of the brain which is valuable for us. So we need to separate the independent signals to obtain the actual physiological signals in order to serve the practical problem. Independent component analysis (ICA) is an effective method for this problem. The main work of this paper is surrounding processing functional magnetic resonance imaging using ICA:1. A new ICA algorithm- new fixed-point (NewFP) is applied to deal with the fMRI data. And the standard information theoretic methods - Akaike s information criterion (AIC) or minimum description length (MDL) is used to estimate the number of the sources and principal component analysis (PCA) is used to reduce the data. These methods can estimate the number of the sources, reduce the superfluous signals and raise the speed of operation. We apply NewFP algorithm and another algorithm (FastICA) to the fMRI data, by comparison, the result shows that NewFP algorithm is superior to the FastICA.2. Three popular methods which can process the signals from a group of subjects- Group independent component analysis (Group ICA) are produced and analyzed, this is very useful for processing fMRI data.3. NewFP algorithm with an useful Group ICA method are used to deal with the fMRI data from group subjects. The result shows that NewFP algorithm is superior to the FastICA. And the result shows using Group ICA to process the data from the group subjects not only can reduce the data, raise the speed of operation but also can ensure the accuracy of the result.更多还原