240 preprocessed images from each subject were concatenated into single four-dimensional images and subjected to ICA using MELODIC (Multivariate Exploratory Linear Decomposition into Independent Components) Version ln(11), part of FSL (FMRIB's Software Library, www.fmrib.ox.ac.uk/fsl). MELODIC implements a Probabilistic ICA estimation (Beckmann and Smith 2004). Following masking of non-brain voxels, voxel-wise de-meaning and normalization of voxel-wise variance, data were whitened and projected into a multi-dimensional subspace using probabilistic Principal Component Analysis. The 4-dimensional data were decomposed into a set of timecourses and spatial maps by optimizing for non-Gaussian spatial source distributions using a fixed-point iteration technique (Hyvarinen 1999). Estimated components were divided by the standard deviation of the residual noise and thresholded by fitting a mixture model to the histogram of intensity values (Beckmann and Smith 2004). The resultant components are comprised of a spatial map of voxels, each assigned a z-score based on likelihood of belonging in the given component. Some components in this analysis represent artifact arising from motion, blood flow or other sources, while other components represent neural networks.
