Reproducibility of graph metrics of human brain functional networks
Reproducibility of graph metrics of human brain functional networks
复制标题
DOI：
10.1016/j.neuroimage.2009.05.035
复制
发表时间：
2009-10-01
期刊：
影响因子：
5.7
通讯作者：
Bassett, Danielle S.
中科院分区：
文献类型：
--
作者：
Deuker, Lorena;Bullmore, Edward T.;Bassett, Danielle S.
图论提供了许多复杂网络组织的度量，可以应用于分析来自神经成像数据的大脑网络。本文研究了16名健康志愿者的脑磁图（MEG）数据的功能网络图指标的重测信度。这些志愿者分别在休息和n-back工作记忆任务中进行了两次测试。对于每个受试者在每次会话中的数据，我们使用小波滤波器来估计在1-60 Hz的总范围内从gamma到低6的每个经典频率区间内每对MEG传感器之间的互信息（MI）。通过阈值化MI矩阵生成无向二值图，并估计了8个全局网络指标：聚类系数、路径长度、小世界性、效率、成本效率、分类性、层次性和同步性。使用类内相关性（ICC）评估每个图度量的可靠性。对于应用于n-back数据的大多数指标，证明了良好的可靠性（平均ICC=0.62）。在低频网络中，可靠性指标更高。高频的伽玛和β频带网络在全球水平上的可靠性较低，但在额叶和顶叶区域显示出高可靠性的节点指标。n-back任务的表现比静息状态数据的测量具有更高的可靠性。任务实践也与更高的可靠性相关。总的来说，这些结果表明，图形度量足够可靠，可以考虑用于未来的功能性脑网络变化的纵向研究。(C) 2009爱思唯尔公司版权所有。
Graph theory provides many metrics of complex network organization that can be applied to analysis of brain networks derived from neuroimaging data. Here we investigated the test-retest reliability of graph metrics of functional networks derived from magnetoencephalography (MEG) data recorded in two sessions from 16 healthy volunteers who were studied at rest and during performance of the n-back working memory task in each session. For each subject's data at each session, we used a wavelet filter to estimate the mutual information (MI) between each pair of MEG sensors in each of the classical frequency intervals from gamma to low 6 in the overall range 1-60 Hz. Undirected binary graphs were generated by thresholding the MI matrix and 8 global network metrics were estimated: the clustering coefficient, path length, small-worldness, efficiency, cost-efficiency, assortativity, hierarchy, and synchronizability. Reliability of each graph metric was assessed using the intraclass correlation (ICC). Good reliability was demonstrated for most metrics applied to the n-back data (mean ICC=0.62). Reliability was greater for metrics in lower frequency networks. Higher frequency gamma and beta-band networks were less reliable at a global level but demonstrated high reliability of nodal metrics in frontal and parietal regions. Performance of the n-back task was associated with greater reliability than measurements on resting state data. Task practice was also associated with greater reliability. Collectively these results suggest that graph metrics are sufficiently reliable to be considered for future longitudinal studies of functional brain network changes. (C) 2009 Elsevier Inc. All rights reserved.