Reproducibility of functional connectivity metrics estimated from resting-state functional MRI with differences in days, coils, and global signal regression
Reproducibility of functional connectivity metrics estimated from resting-state functional MRI with differences in days, coils, and global signal regression
复制标题
DOI：
10.1007/s12194-022-00670-6
复制
发表时间：
2022-08-12
影响因子：
1.6
通讯作者：
Sobue，Gen
中科院分区：
文献类型：
--
作者：
Kato，Sanae;Bagarinao，Epifanio;Sobue，Gen
在多中心研究中，使用静息态功能磁共振成像（rs-fMRI）检查脑功能变化时，成像采集系统的差异可能会影响结果的可重复性。这对于纵向研究也很重要，因为设备设置可能会发生变化。本研究探讨了功能连接（FC）的指标估计的再现性，从使用扫描仪接收线圈与不同数量的通道获得的rs-fMRI数据。这项研究涉及来自20名健康志愿者的80个rs-fMRI数据集，每个数据集使用12通道和32通道线圈进行两次独立的成像扫描。我们使用独立成分分析（伊卡）来评估典型的静息态网络（RSN）的FC和图论来计算几个全脑网络度量。还考虑了全局信号回归（GSR）作为预处理步骤的效果。进行了接收器线圈内和之间的比较。无论GSR，RSN来自rs-fMRI数据采集使用相同的接收器线圈是可重现的，但不是从不同的接收器线圈。然而，GSR和接收器线圈的通道计数对使用全脑网络分析估计的网络度量的再现性具有可辨别的影响。使用32通道线圈采集的数据往往比使用12通道线圈采集的数据具有更好的再现性。我们的研究结果表明，使用不同的接收器线圈获得的RS-fMRI数据估计的FC指标的再现性表现出一定程度的依赖于预处理方法和进行的分析类型。
In multisite studies, differences in imaging acquisition systems could affect the reproducibility of the results when examining changes in brain function using resting-state functional magnetic resonance imaging (rs-fMRI). This is also important for longitudinal studies, in which changes in equipment settings can occur. This study examined the reproducibility of functional connectivity (FC) metrics estimated from rs-fMRI data acquired using scanner receiver coils with different numbers of channels. This study involved 80 rs-fMRI datasets from 20 healthy volunteers scanned in two independent imaging sessions using both 12- and 32-channel coils for each session. We used independent component analysis (ICA) to evaluate the FC of canonical resting-state networks (RSNs) and graph theory to calculate several whole-brain network metrics. The effect of global signal regression (GSR) as a preprocessing step was also considered. Comparisons within and between receiver coils were performed. Irrespective of the GSR, RSNs derived from rs-fMRI data acquired using the same receiver coil were reproducible, but not from different receiver coils. However, both the GSR and the channel count of the receiver coil have discernible effects on the reproducibility of network metrics estimated using whole-brain network analysis. The data acquired using the 32-channel coil tended to have better reproducibility than those acquired using the 12-channel coil. Our findings suggest that the reproducibility of FC metrics estimated from rs-fMRI data acquired using different receiver coils showed some level of dependence on the preprocessing method and the type of analysis performed.