Estimating Components of Functional Magnetic Resonance Imaging (Fmri) Data in a Time-Dependent Order by Modifying Ica Algorithms (original) (raw)

In this paper, our aim is analyzing functional magnetic resonance imaging (fMRI) data by independent component analysis (ICA) in order to find regions of brain which were activated by neural activity in human brain. Usually by applying ICA algorithm for whole dataset, independent components can be estimated but we can't understand the procedure of activation. Here, we propose a method to detect active components in different time intervals. Spatial ICA is applied in sliding time windows. We find active components in each window by applying a criteria which measure two kind of cross-correlation coefficients: the correlation between components in each window and reference function in that time interval and the correlation between components in adjacent windows. Finally, we detect active regions of active components in each window. In order to investigate the advantage of using this method, we perform some experiments for simulated and experimental fMRI datasets and show the results. Receiver operating characteristic (ROC) curve shows the performance of this method.