MEG source reconstruction constrained by diffusion MRI based whole brain dynamical model

Abstract

Previous studies have shown that MEG source reconstruction is improved by temporal constraints from local current source dynamics. Extending these constraints, we have developed a source reconstruction method that is spatiotemporally constrained by a whole brain dynamical model. The source dynamics are represented by a multivariate autoregressive (MAR) model whose matrix entries are constrained by connectivity estimates based on diffusion MRI. The MAR model parameters are jointly estimated with the source amplitude to infer source-space effective connectivity. Through simulation at low signal-to-noise ratio, we confirmed that the proposed method suppresses spurious sources and, unlike the non-dynamical sparse Bayesian method, can recover a low amplitude source. Furthermore, effective connectivity estimated by the proposed joint approach was more accurate than that obtained from the two stage approach, in which the current sources are first reconstructed by the non-dynamical method, followed by MAR model fitting to the resulting sources.