Multisensory integration involves multiple cortical regions and occurs at multiple stages with attentional modulation. The structure of network formed by the interactive cortical regions reflects the state of working on a current task and changes continuously with the task processing. In addition, the neural oscillatory responses in various frequency bands are associated with different cognitive functions. Thus, studying topological characteristics of time-varying networks across multiple frequency bands helps to elucidate the mechanism of multisensory integration. Here, we designed an event-related experiment using auditory, visual, and audiovisual stimuli to record electroencephalographic data in both attended and unattended conditions and constructed delta-, theta-, alpha-, and beta-band networks at eight time points post-stimulus. We used graph theory to calculate global properties, nodal out-degree, and their correlation with behavioral performance. The increasing clustering coefficient and global efficiency and decreasing characteristic path length indicated that the brain had optimized the configuration across multiple frequency bands over time to efficiently process audiovisual integration. The differences in global properties and hub distributions showed that each frequency band–specificity system in the brain had a different topological structure, indicating that the networks on each frequency band contributed to various cognitive functions and involved in different stages of audiovisual integration. Our results suggest that differences in cognitive function are, at least partly, due to the different network structures across frequency bands and that the frequency band–specificity systems with different distribution are involved in various stages of audiovisual integration and attention modulation.

多感官整合涉及多个皮层区域，并在注意力调节的多个阶段发生。交互皮质区域形成的网络结构反映了当前任务的工作状态，并随着任务处理而不断变化。另外，各个频带中的神经振荡响应与不同的认知功能相关。因此，研究跨越多个频带的时变网络的拓扑特性有助于阐明多感官整合的机制。在这里，我们设计了一个使用听觉，视觉和视听刺激的事件相关实验，以记录在有人和无人照管情况下的脑电图数据，并在刺激后的八个时间点构建了δ，θ，α和β带网络。我们使用图论来计算全局属性，节点向外度及其与行为表现的相关性。聚类系数的增加和全局效率的提高以及特征路径长度的减小表明，随着时间的流逝，大脑已经跨多个频段优化了配置，以有效地处理视听集成。全局属性和集线器分布的差异表明，大脑中的每个频带特异性系统都具有不同的拓扑结构，这表明每个频带上的网络有助于各种认知功能，并参与视听整合的不同阶段。我们的结果表明，认知功能的差异至少部分是，