How reliable are the functional connectivity networks 1 of MEG in resting states ?

19 We investigated the reliability of nodal network metrics of functional connectivity (FC) 20 networks of MEG covering the whole brain at the sensor level in the eyes-closed (EC) and 21 eyes-open (EO) resting states. Mutual information (MI) was employed as a measure of FC 22 between sensors in theta, alpha, beta, and gamma frequency bands of MEG signals. MI matrices 23 were assessed with three nodal network metrics, i.e., nodal degree (Dnodal), nodal efficiency 24 (Enodal), and betweenness centrality (normBC). Intra-class correlation (ICC) values were 25 calculated as a measure of reliability. We observed that the test-retest reliabilities of the resting 26 states ranged from a poor to good level depending on the bands and metrics used for defining the 27 nodal centrality. The dominant alpha band FC network changes were the salient features of the 28 state-related FC changes. The FC networks in the EO resting state showed greater reliability 29 when assessed by Dnodal (maximum mean ICC=0.655) and Enodal (maximum mean 30 ICC=0.604) metrics. The gamma band FC network was less reliable than the theta, alpha, and 31 beta networks across the nodal network metrics. However, the sensor-wise ICC values for the 32 nodal centrality metrics were not uniformly distributed, that is, some sensors had high reliability. 33 This study provides a sense of how the nodal centralities of the human resting state MEG are 34 distributed at the sensor level and how reliable they are. It also provides a fundamental scientific 35 background for continued examination of the resting state of human MEG.