Independent component analysis of fMRI data: Examining the assumptions

摘要: Independent component analysis (ICA), which separates fMRI data into spatially independent patterns of activity, has recently been shown to be a suitable method for exploratory analysis. The validity the assumptions ICA, mainly that underlying components are and add linearly, was explored with representative set by calculating log-likelihood observing each voxel's time course conditioned on ICA model. probability courses from white-matter voxels higher compared other observed brain regions. Regions containing blood vessels had lowest probabilities. statistical distribution probabilities over all did not resemble expected small number mixed Gaussian noise. These results suggest model may more accurately represent in specific regions brain, both activity-dependent sources flow noise non-Gaussian.