A0658
Title: Cyclic causality: Unraveling nonlinear interactions in brain time series data
Authors: Hernando Ombao - King Abdullah University of Science and Technology (KAUST) (Saudi Arabia)
Moo K Chung - University of Wisconsin-Madison (United States) [presenting]
Anass El Yaagoubi Bourakna - King Abdullah University of Science and Technology (Saudi Arabia)
Abstract: A novel framework is proposed for detecting and quantifying cyclic causality in brain time series data using the Hodge decomposition, with the aim of uncovering nonlinear and higher-order interactions often missed by conventional linear models such as Granger causality. Traditional causal models assume acyclic dependencies, making them inadequate for modeling the recurrent, feedback-driven dynamics commonly found in resting-state functional human brain networks. The approach represents time-varying brain connectivity as functions over simplicial complexes and leverages the Hodge Laplacian to extract topological structures, particularly 1-cycles, that reflect causal loops in the network. Persistent homology is used to assess the stability of these features across time and subjects. Spectral Hodge expansions are further introduced to construct topological time-frequency representations, or topological spectrograms, allowing for fine-grained dynamic analysis.
