新的独立成分分析算法结合主成分分析实现fMRI信号的盲分离

Blind Source Separation for fMRI Signals Using a New Independent Component Analysis Algorithm and Principal Component Analysis

用ICA算法来实现fMRI信号的盲源分离,可以提取出产生fMRI信号的多种源信号。但是在处理过程中存在两个困难:(1)fMRI数据的规模比较大,计算耗时;(2)计算量太大难免产生误差,给结果的分析带来不便。所以我们考虑对数据进行降维,但是如何确定源信号的个数也是一个难题。我们利用信息论的方法来估计源信号的个数,再使用主成分分析对数据进行降维。通过这样的处理,有效地确定了源信号的个数,减少了计算量。然后将一种新的ICA算法(New fixed-point,NewFP)用于处理降维后的数据。最后通过对实际的fMRI信号进行处理,结果表明新算法可以快速有效的分离fMRI信号,且准确性优于FastICA算法。

The application of independent component analysis （ICA） to the functional magnetic resonance imaging （fMRI） data can separate many independent sources. But in the processing there are two difficulties： （1） the data of the fMRI is usually on a large scale, so the computing is time-consuming; （2） we cannot avoid the errors for too heavy computational load, this brings many troubles. Thus we think of reducing the data. In this article we used the standard information theoretic methods to estimate the number of the sources and used the principal component analysis （PCA） to reduce the data. By this process, we estimated the number of the sources and reduced the data successfully; Then we applied the ICA algorithm to the reduced fMRI data; this method raised the speed of operation. After application of the new ICA algorithm and another algorithm （FastlCA） to the fMRI data, a comparison was made. The results show that the new algorithm can separate the fMRI data fast and effectively and it is superior to the FastICA on the accuracy of estimating the temporal dynamics of activations.