This article introduces the measure of integrated local correlation (ILC) for assessing local coherence in the brain using functional magnetic resonance imaging (fMRI) data and characterizes the measure in terms of reproducibility, the effect of physiological noise, and the dependence on image resolution. The coupling of local neuronal processes influences coherence in a voxel's neighborhood. ILC is defined, for each voxel, as the integration of its spatial correlation function. This integrated measure does not require the specification of a neighborhood and, as demonstrated by experimental data, is effectively independent of image resolution. Respiratory and cardiac fluctuations do not considerably alter the ILC value except in isolated areas in and surrounding large vessels. With resting-state fMRI data, ILC was demonstrated to be tissue-specific, higher in gray matter than white matter, and reproducible across consecutive runs in healthy individuals. Within the gray matter, ILC was found to be higher in the default mode network, particularly the posterior and anterior cingulate cortices. Comparing ILC maps obtained from resting state and continuous motor task data, we observed reduced local coherence in the default mode network during the task. Finally, we compared ILC and regional homogeneity by examining their ability to discriminate between gray and white matters in resting state data and found ILC to be more sensitive.
(c) 2007 Wiley-Liss, Inc.