基于改进的动态因果模型量化生理模型的效应连通性

his paper proposes a new strategy named Modified Dynamic Causal Modelling (MDCM) to detect effective connectivity in an underlying physiology based model aimed at simulating epileptic activities in the hippocampus. The MDCM consists of three parts considering a state-space environment. First of all, an analytical form of the power spectral densities (PSD) as a function of the model parameters is obtained by calculating the transition matrix of the state-space linearized model. Secondly, the PSD are measured from the observed intracerebral electroencephalographic (iEEG) signals, and then the model parameters are estimated by an Expectation Maximum (EM) algorithm. Thirdly, a new estimation of the parameters is obtained using an iterative algorithm and another estimation of the PSD is derived, the most plausible model being finally selected among a set of candidate models. Simulated experimental results indicate that the proposed MDCM outperforms DCM in scenarios of unidirectional flow propagation.