Abstract
In this paper, we present a framework to reconstruct spatially localized sources from Magnetoencephalography (MEG)/Electroencephalography (EEG) using spatiotemporal constraint. The source dynamics are represented by a Multivariate Autoregressive (MAR) model whose matrix elements are constrained by the anatomical connectivity obtained from diffusion Magnetic Resonance Imaging (dMRI). The framework assumes that the whole brain dynamic follows a constant MAR model in a time window of interest. The source activations and the MAR model parameters are estimated iteratively. We could confirm the accuracy of the framework using simulation experiments in both high and low noise levels. The proposed framework outperforms the two-stage approach.
Highlights
EEG and MEG are two non-invasive modalities that allow us to measure brain activity with high temporal resolution
We provide more degrees of freedom by allowing source interactions constrained by anatomical connectivity obtained from diffusion Magnetic Resonance Imaging (dMRI) information
Through simulations at different noise levels, we compare the accuracy of the proposed framework in the source reconstruction and multivariate autoregressive (MAR) model parameters estimation with the ones obtained from a non-dynamical sparse solution based on mixed-norm estimates (MxNE) [7]
Summary
EEG and MEG are two non-invasive modalities that allow us to measure brain activity with high temporal resolution. Other studies, [4], [5], assume that sources increase or decrease by the same factor or interact with only their direct neighbors These assumptions significantly reduce the degrees of freedom of the dynamics. We provide more degrees of freedom by allowing source interactions constrained by anatomical connectivity obtained from dMRI information. This allows us to generalize our work to the whole brain. The source dynamics represented by the MAR model is based on the underlying anatomical connections obtained from dMRI information. Through simulations at different noise levels, we compare the accuracy of the proposed framework in the source reconstruction and MAR model parameters estimation with the ones obtained from a non-dynamical sparse solution based on mixed-norm estimates (MxNE) [7]. Our framework uses the MxNE estimates to initialize the source intensities estimates the MAR entries iteratively
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
