Application Of Independent Component Analysis In FMRI Data Processing

The brain is extremely complex product.For many years, human has been trying to interpret the brain——"black box". With the development of brain imaging technology, the research on the brain has transferred from the anatomical localization at the beginning to the in-depth probe in the basic process of functional activities of the brain. Functional magnetic resonance imaging (fMRI) is a kind of imaging technique, developed in the1990s, based on the blood oxygen level dependent and blood flow sensitive. Functional Magnetic Resonance Imaging, as a non-invasive technique, has its own advantages rather than other techniques. Therefore, in recent years, it has become the first choice of studying brain activities. Correspondingly, many a method of analysis of fMRI data has appeared. In this paper, a novel independent component analysis is introduced and its application in functional magnetic resonance imaging data.In the past, methods for processing fMRI data require priori assumption about the time series according to experimental design and analysis of the correlation of brain active regions. If time series collected by the brain is unknown, factors that activate brain regions and its impact are also unknown, and then it means that the signal source and the mixing matrix is unknown, which is a typical of blind source separation problem. To solve the problem of blind source separation, we can choose the most popular method independent component analysis, through which the fMRI data is separated and task-related components are extracted.This paper first introduces the fMRI theory, applications, features, and independent component analysis with some estimated algorithm; a novel data analysis method is also introduced, which is based on Kurtosis RobustICA (Robust Independent Component Analysis). And its theory, the steps and advantage are analyzed. A number of simulation experiments show that the algorithm has been greatly improved over the previous FastICA algorithm; the materials and methods of the experiment are briefly introduced; fMRI data is separated by independent component analysis (RobustICA algorithm and the FastICA algorithm). Compared with the analysis of the FastICA algorithm, because of RobustICA algorithm robustness being better than FastICA algorithm,21independent components are successfully separated by RobustICA algorithm, while19independent components are separated by FastICA algorithm. The other two independent components are task-related component and noise component, which FastICA algorithm fails to separate.