A mechanistic whole brain model to capture simultaneous EEG-fMRI data

This study introduces a novel oscillatory network model to simulate simultaneous electroencephalography-functional magnetic resonance (EEG-fMRI) data, addressing the reconstruction challenge that arises due to their contrasting spatiotemporal scales. Here, each brain region is modeled by 2 oscillator clusters-a cluster of low-frequency (LFO) and high-frequency Hopf oscillators (HFO) coupled with an innovative power-coupling rule, facilitating cross-frequency interactions. The model is trained in 2 stages: learning oscillators' frequencies and phase relations using a biologically plausible complex-Hebbian rule in the first stage and, followed by a modified complex backpropagation for amplitude approximation, overcoming limitations of poor accuracy and computational complexity in existing models. This framework outperforms current methods in replicating empirical functional connectivity (FC), FC dynamics (FCD), and modularity over disparate spatio-temporal scales. The correlation between the FC of fMRI and the FCs of various EEG frequency bands is reflected in the strengths of the LFO-HFO coupling. Furthermore, in silico structural perturbation studies quantified the effect of pruning of the anatomical connectivity on spatiotemporal dynamics in terms of FC, FCD, modularity, and integration level integrated state of occurrence rate. The model's ability to reconstruct simultaneous EEG-fMRI data showcases significant advancement in understanding the resting-state brain's functionality from multimodal settings and deciphering neurological disorders in diverse spatiotemporal scales.