Nalci Alican, Rao Bhaskar D, Liu Thomas T
Center for Functional MRI, University of California San Diego, 9500 Gilman Drive MC 0677, La Jolla, CA 92093, United States; Department of Electrical and Computer Engineering, University of California San Diego, 9500 Gilman Drive, La Jolla, CA 92093, United States.
Department of Electrical and Computer Engineering, University of California San Diego, 9500 Gilman Drive, La Jolla, CA 92093, United States.
Neuroimage. 2017 May 15;152:602-618. doi: 10.1016/j.neuroimage.2017.01.015. Epub 2017 Jan 9.
In resting-state functional MRI (rsfMRI), the correlation between blood oxygenation level dependent (BOLD) signals across different brain regions is used to estimate the functional connectivity of the brain. This approach has led to the identification of a number of resting-state networks, including the default mode network (DMN) and the task positive network (TPN). Global signal regression (GSR) is a widely used pre-processing step in rsfMRI that has been shown to improve the spatial specificity of the estimated resting-state networks. In GSR, a whole brain average time series, known as the global signal (GS), is regressed out of each voxel time series prior to the computation of the correlations. However, the use of GSR is controversial because it can introduce artifactual negative correlations. For example, it has been argued that anticorrelations observed between the DMN and TPN are primarily an artifact of GSR. Despite the concerns about GSR, there is currently no consensus regarding its use. In this paper, we introduce a new framework for understanding the effects of GSR. In particular, we show that the main effects of GSR can be well approximated as a temporal downweighting process in which the data from time points with relatively large GS magnitudes are greatly attenuated while data from time points with relatively small GS magnitudes are largely unaffected. Furthermore, we show that a limiting case of this downweighting process in which data from time points with large GS magnitudes are censored can also approximate the effects of GSR. In other words, the correlation maps obtained after GSR show a high degree of spatial similarity (including the presence of anticorrelations between the DMN and TPN) with maps obtained using only the uncensored (i.e. retained) time points. Since the data from these retained time points are unaffected by the censoring process, this finding suggests that the observed anticorrelations inherently exist in the data from time points with small GS magnitudes and are not simply an artifact of GSR.
在静息态功能磁共振成像（rsfMRI）中，不同脑区之间的血氧水平依赖（BOLD）信号相关性被用于估计大脑的功能连接性。这种方法已导致识别出许多静息态网络，包括默认模式网络（DMN）和任务积极网络（TPN）。全局信号回归（GSR）是rsfMRI中广泛使用的预处理步骤，已被证明可提高估计的静息态网络的空间特异性。在GSR中，在计算相关性之前，将全脑平均时间序列（称为全局信号，GS）从每个体素时间序列中回归出去。然而，GSR的使用存在争议，因为它可能会引入人为的负相关性。例如，有人认为在DMN和TPN之间观察到的反相关性主要是GSR的伪影。尽管对GSR存在担忧，但目前对于其使用尚无共识。在本文中，我们引入了一个用于理解GSR效果的新框架。特别是，我们表明GSR的主要效果可以很好地近似为一个时间加权过程，其中来自GS幅度相对较大时间点的数据被大幅衰减，而来自GS幅度相对较小时间点的数据基本不受影响。此外，我们表明这种加权过程的一个极限情况，即对GS幅度较大时间点的数据进行审查，也可以近似GSR的效果。换句话说，GSR后获得的相关图与仅使用未审查（即保留）时间点获得的图显示出高度的空间相似性（包括DMN和TPN之间存在反相关性）。由于来自这些保留时间点的数据不受审查过程的影响，这一发现表明观察到的反相关性在GS幅度较小时间点的数据中固有存在，而不仅仅是GSR的伪影。
