Huster René J, Raud Liisa
Multimodal Imaging and Cognitive Control Lab, Department of Psychology, University of Oslo, Oslo, Norway.
Psychology Clinical Neurosciences Center, University of New Mexico, Albuquerque, USA.
Brain Topogr. 2018 Jan;31(1):3-16. doi: 10.1007/s10548-017-0603-x. Epub 2017 Oct 23.
Over the last years we saw a steady increase in the relevance of big neuroscience data sets, and with it grew the need for analysis tools capable of handling such large data sets while simultaneously extracting properties of brain activity that generalize across subjects. For functional magnetic resonance imaging, multi-subject or group-level independent component analysis provided a data-driven approach to extract intrinsic functional networks, such as the default mode network. Meanwhile, this methodological framework has been adapted for the analysis of electroencephalography (EEG) data. Here, we provide an overview of the currently available approaches for multi-subject data decomposition as applied to EEG, and highlight the characteristics of EEG that warrant special consideration. We further illustrate the importance of matching one's choice of method to the data characteristics at hand by guiding the reader through a set of simulations. In sum, algorithms for group-level decomposition of EEG provide an innovative and powerful tool to study the richness of functional brain networks in multi-subject EEG data sets.
在过去几年中，我们看到大型神经科学数据集的相关性稳步增加，随之而来的是对分析工具的需求也在增长，这些工具需要能够处理如此大规模的数据集，同时提取跨受试者具有普遍性的大脑活动特性。对于功能磁共振成像，多受试者或组水平的独立成分分析提供了一种数据驱动的方法来提取内在功能网络，如默认模式网络。与此同时，这种方法框架已被应用于脑电图（EEG）数据的分析。在这里，我们概述了目前可用于EEG多受试者数据分解的方法，并强调了EEG中值得特别考虑的特征。我们还通过一组模拟引导读者，进一步说明了根据手头数据特征选择合适方法的重要性。总之，EEG组水平分解算法为研究多受试者EEG数据集中功能脑网络的丰富性提供了一种创新且强大的工具。
