Abstract
Electro/Magnetoencephalography (EEG/MEG) source-space network analysis is increasingly recognized as a powerful tool to track fast electrophysiological brain dynamics. However, an objective and quantitative evaluation of the various steps, from source localization and functional connectivity to clustering algorithms, is challenging, due to the lack of realistic ‘controlled’ data. Here, we used a human brain computational model containing both physiologically-based cellular GABAergic and Glutamatergic circuits coupled through Diffusion Tensor Imaging - based structural connectivity, to generate realistic High Density-EEG (256 channels) recordings. We designed a scenario of successive gamma-band oscillations in distinct cortical areas in order to emulate a virtual picture naming task. We identified the fast time-varying network states and quantified the performance of the key steps involved in the pipeline: (1) inverse models to reconstruct cortical-level sources, (2) functional connectivity measures to compute statistical interdependency between regional time series, and (3) dimensionality reduction methods to derive dominant brain network states (BNS). Using a systematic evaluation of the different independent/principal/non-negative decomposition techniques along with a clustering approach, results show significant variability among the tested algorithms in terms of spatial and temporal accuracy. We outlined the spatial precision, the temporal sensitivity, as well as the global accuracy of the extracted BNS relative to each method. Our findings suggest a good performance of wMNE/PLV combination to elucidate the appropriate functional networks and ICA techniques to derive relevant dynamic brain network states. Our aim here is twofold: 1) to provide quantitative assessment on the advantages and the limitations of each of the analyzed techniques and 2) to introduce (and share) a complete framework that can be used to optimize the entire pipeline of EEG/MEG source connectivity. With such framework, other tasks can be generated and used for validation and other methodological points can be also addressed.
Competing Interest Statement
The authors have declared no competing interest.