Yang Ruochen, Ping Heng, Xiao Xiongye, Kiani Roozbeh, Bogdan Paul
Ming Hsieh Department of Electrical and Computer Engineering, University of Southern California, Los Angeles, CA, USA.
Min H. Kao Department of Electrical Engineering and Computer Science, University of Tennessee, Knoxville (UTK), Knoxville, TN, USA.
Nat Commun. 2025 Jul 30;16(1):6994. doi: 10.1038/s41467-025-62202-1.
Brain networks exhibit diverse topological structures to adapt and support brain functions. The changes in neuronal network architecture can lead to alterations in neuronal spiking activity, yet how individual neuronal behavior reflects network structure remains unexplored. Therefore, mathematical tools to decode and infer neuronal network structure and role from spiking behavior need to be developed to relate the neuronal firing activity with topology and goal of underlying network. Toward this end, we perform a comprehensive multifractal analysis of the neuronal interspike intervals to characterize their non-linear, non-stationary and non-Markovian dynamics. We explore the relationship of neuronal network connectivity with the multifractal spiking pattern and show that such a measure is sensitive to network structure while relatively consistent to stimulus. In addition, we reveal that the observed multifractal profile is not influenced by the activity of unobserved neuronal ensembles. To mimic neurons performing specific functions, we further train spiking neural networks to generate goal-directed architectures and demonstrate that multifractal analysis also enables differentiating networks with diverse tasks.
脑网络呈现出多样的拓扑结构以适应和支持脑功能。神经元网络结构的变化会导致神经元放电活动的改变，然而个体神经元行为如何反映网络结构仍未得到探索。因此，需要开发数学工具来从放电行为中解码和推断神经元网络结构及作用，以便将神经元放电活动与基础网络的拓扑结构和目标联系起来。为此，我们对神经元峰峰间期进行了全面的多重分形分析，以表征其非线性、非平稳和非马尔可夫动力学。我们探索了神经元网络连通性与多重分形放电模式之间的关系，结果表明这种度量对网络结构敏感，而对刺激相对一致。此外，我们发现观察到的多重分形谱不受未观察到的神经元集合活动的影响。为了模拟执行特定功能的神经元，我们进一步训练脉冲神经网络以生成目标导向的架构，并证明多重分形分析也能够区分执行不同任务的网络。
