Correlations in spontaneous bold (blood oxygen level dependent) fluctuations differ depending on resting state condition

A number of brain networks can be identified by measuring the inter-regional temporal correlation of slow (< 0.1 Hz) spontaneous, i.e. not task-evoked, fluctuations of the BOLD signal. Although several studies have reported similar topography of these networks across different resting conditions, sleep, and anesthesia, differences between resting conditions have received little attention. Here, we compare the inter-regional correlation structure within different networks and between networks across different resting conditions. Ten normal subjects underwent three 5.5-min MRI (magnetic resonance imaging) scans in three different conditions: visual fixation on a crosshair (FIX), eyes closed (EC), and eyes open (EO) in low-level illumination (without fixation) (Fox et al., 2005). We analyzed the temporal correlation of BOLD signals between different retinotopic visual areas (V1, V2, V3, V3A, V4, V7, VP) and other brain networks: dorsal and ventral attention, default, language, and motor. We also analyzed the temporal correlation of the signal between regions within each network. As previously reported (Fox et al., 2005), we found weak or no effect of different behavioral states on inter-regional correlations within each network, except for the visual system where the signal correlation decreased when going from EC to EO to FIX. In contrast, we found highly specific effects of behavioral state on connectivity between networks. Right frontal eye field (FEF) was more strongly correlated with V7 bilaterally during FIX and EO than EC. This difference was not observed in left FEF or between right FEF and any other region outside of the visual system, and was stronger in FEF than in other regions of the dorsal attention network (posterior IPS (intra-parietal sulcus) and ventral IPS). The linkage of right FEF with a higher-order visual area during fixation and eyes open is consistent with the putative role of this region as a source of top-down control on visual cortex. Secondly, we observed a very different pattern of correlation between primary motor cortex and V7, with strong correlation during EC, intermediate correlation during FIX, and no correlation during EO. These findings show that different networks become correlated in different behavioral states while preserving internally a high degree of temporal coherence.