Multi-group semi-blind ICA of fMRI

点击翻译按钮获取中文摘要

DESCRIPTION (provided by applicant): The overall goal of this proposed research is to develop, evaluate the performance of, and optimize semi-blind independent component analysis (sblCA) methods for the analysis of multi-group (or multi-paradigm) functional magnetic resonance imaging (fMRI) data. We will examine the application of ICA to fMRI of visual, auditory, and motor paradigms using a proposed synthesis/analysis model [1,2] (and subsequently developed extensions of it) to carefully study the properties of ICA as it is applied to fMRI data and incorporate additional a priori information. We recently published a method for applying independent component analysis (ICA) to groups of subjects [3,4]. We will extend this method to enable comparisons of multiple groups. The use of flexible methods, such as ICA, is of _articular importance when studying cognitive tasks involving a distributed set of brain regions or when studying a disease like schizophrenia, which is known to be a diffuse disorder, affecting many aspects of brain function. We will apply our methods to reanalyze data, previously analyzed with traditional analysis methods, acquired while normal and schizophrenic subjects performed an auditory oddball paradigm. The specific aims of the proposed research are 1) To develop methods for inter-group/inter-paradigm inference which a) incorporate a priori information about the brain sources, b) provide the probability of source magnitude across and between subjects, c) utilize post hoc parameterization to obtain, e.g., latency valued, and d) enable hypothesis testing when a priori time courses or activation locations are predicted. 2) To optimize and validate ICA of fMRI using "basic" visual/motor/auditory paradigms, realistic simulations using our model for ICA of fMRI, "hybrid" data sets containing known sources, and comparisons with SPM99 methods assuming a known form for the hemodynamic response. 3) To apply our methods to analyze previously acquired and well characterized data of normal and schizophrenic subjects performing an auditory oddball task, and 4) To make publicly available, a Matlab software package implementing our methods which can be used stand alone or as a plug-in for a widely used fMRI analysis package, SPM.