[[SUPERSEDED - this dataset is replaced by a later version: https://doi.org/10.7488/ds/2109]] Accounting for the Complex Hierarchical Topology of EEG Functional Connectivity in Network Binarisation

[[SUPERSEDED - this dataset is replaced by a later version: https://doi.org/10.7488/ds/2109]] Research into binary network analysis of brain function faces a methodological challenge in selecting an appropriate threshold to binarise edge weights. For EEG, such binarisation should take into account the complex hierarchical structure found in functional connectivity. We explore the density range suitable for such structure and provide a comparison of state-of-the-art binarisation techniques, the recently proposed Cluster-Span Threshold (CST), minimum spanning trees and union of shortest path graphs, with arbitrary proportional thresholds and weighted networks. We test these techniques on weighted complex hierarchy models by contrasting model realisations with small parametric differences. We also test the robustness of these techniques to random and targeted topological attacks. We reveal that complex hierarchical topology requires a medium-density range binarisation solution, such as the CST which proves near maximal for distinguishing differences when compared with arbitrary proportional thresholding. Simulated results are validated with the analysis of three relevant EEG datasets: eyes open and closed resting states; visual short-term memory tasks; and resting state Alzheimer's disease with a healthy control group. The CST consistently outperforms other state-of-the-art binarisation methods for topological accuracy and robustness in both synthetic and real data. We provide insights into how the complex hierarchical structure of functional networks is best revealed in medium density ranges and how it safeguards against targeted attacks. These EEG PLI connnectivity data sets are used in the analysis of our submitted manuscript: https://arxiv.org/abs/1610.06360.

【已废弃——本数据集已被后续版本替换：https://doi.org/10.7488/ds/2109】脑功能二值网络分析领域的研究面临一项方法学挑战：如何选取合适的阈值对边权重进行二值化处理。脑电图（Electroencephalogram, EEG）的二值化处理需要考虑功能连接中存在的复杂层级结构。本研究探索了适配该层级结构的密度区间，并对当前主流的二值化技术进行了对比分析，包括新近提出的簇跨度阈值（Cluster-Span Threshold, CST）、最小生成树、最短路径图并集，以及任意比例阈值法与加权网络法。我们通过对比参数存在细微差异的模型实现结果，在加权复杂层级模型上对这些技术进行了测试。此外，我们还测试了这些技术在随机拓扑攻击与针对性拓扑攻击下的鲁棒性。研究表明，复杂层级拓扑结构需要采用中等密度区间的二值化方案，例如CST，相较于任意比例阈值法，该方法在区分差异时的表现接近最优。我们通过分析三个相关脑电图数据集对仿真结果进行了验证：睁眼与闭眼静息态数据集、视觉短时记忆任务数据集，以及伴健康对照组的阿尔茨海默病静息态数据集。在合成数据与真实数据中，CST在拓扑准确性与鲁棒性方面始终优于其他当前主流的二值化方法。本研究揭示了功能网络的复杂层级结构在中等密度区间下的最优展现方式，以及该结构如何抵御针对性攻击。本研究所用的脑电图相位滞后指数（Phase Lag Index, PLI）连接数据集已用于我们已投稿的论文分析：https://arxiv.org/abs/1610.06360。