Frequencies contributing to functional connectivity in the cerebral cortexin "resting-state" data

BACKGROUND AND PURPOSE: In subjects performing no specific cognitive task ( "resting state"), time courses of voxels within functionally connected regi ons of the brain have high cross-correlation coefficients ("functional conn ectivity"). The purpose of this study was to measure the contributions of l ow frequencies and physiological noise to cross-correlation maps. METHODS: In four healthy volunteers, task-activation functional MR imaging and resting-state data were acquired. We obtained four contiguous slice loc ations in the "resting state" with a high sampling rate. Regions of interes t consisting of four contiguous voxels were selected. The correlation coeff icient for the averaged time course and every other voxel in the four slice s was calculated and separated into its component frequency contributions. We calculated the relative amounts of the spectrum that were in the low-fre quency (0 to 0.1 Hz), the respiratory-frequency (0.1 to 0.5 Hz), and cardia c-frequency range (0.6 to 1.2 Hz). RESULTS: For each volunteer, resting-state maps that resembled task-activat ion maps were obtained. For the auditory and visual cortices, the correlati on coefficient depended almost exclusively on low frequencies (<0.1 Hz). Fo r all cortical regions studied, low-frequency fluctuations contributed more than 90% of the correlation coefficient. Physiological (respiratory and ca rdiac) noise sources contributed less than 10% to any functional connectivi ty MR imaging map. In blood vessels and cerebrospinal fluid, physiological noise contributed more to the correlation coefficient. CONCLUSION. Functional connectivity in the auditory, visual, and sensorimot or cortices is characterized predominantly by frequencies slower than those in the cardiac and respiratory cycles. In functionally connected regions, these low frequencies are characterized by a high degree of temporal cohere nce.