功能磁共振图像处理的ICA方法综述

Independent Component Analysis of Functional MRI Data: An Overview

能够进行无损伤探测的功能核磁共振成像(functionalmagneticresonanceimaging,fMRI)技术使人们又多了一种研究神经活动的有力工具,但传统的数据分析方法还不能很好地揭示fMRI数据中所包含的丰富信息,而独立分量分析(independentcomponentanalysis,ICA)作为一种新近出现的数据处理方法,则不仅可以从fMRI数据分析中得出一些传统方法所未发现的结果,并且这种方法不需要传统方法的那种预先假设的先验模型,只依赖于数据本身即可提取其中所包含的信息。为了使人们对这一技术有一概略了解,首先对ICA方法的基本原理及其在fMRI数据处理中的应用进行了综述,并针对不同特点的fMRI数据详细讨论了如何选择不同的算法;然后ICA方法与传统方法相比存在的优越性进行了介绍,最后提出了此方法当前存在的一些问题及处理思路,并展望了其在fMRI数据处理中的发展趋势,可以认为,ICA是一种很有发展潜力的功能磁共振数据处理新方法。

Functional magnetic resonance imaging is a non-invasive powerful tool for people to investigate the neuronal activity in vivo. But the abundant information contained in the fMRI data could not be fully mined by the traditional methods. As a new promising approach, independent component analysis revealed some components in fMRI data which can not be detected by the conventional methods. Depending only on data set itself, this exploratory analysis technique doesn’t need the a priori model used by the conventional methods. The primary principle and the application on the fMRI data were reviewed followed by the detailed discussion on the selection of the ICA algorithms. Compared to the traditional methods, the advantages of ICA could be found through the application results from many papers. Some issues were discussed and the trends of this method applied to the fMRI data were proposed by the author in the final part of this paper, which support the standpoint that this new potential approach could be used to mining the fMRI data.