Dynamic state allocation for MEG source reconstruction

• DOI: 10.1016/j.neuroimage.2013.03.036
• Publication Date: 2013-03-29T13:32:05Z
• Authors: Mark W. Woolrich; Adam Baker; Henry Luckhoo; Hamid Mohseni; Gareth Barnes; Matthew Brookes; Iead Rezek
• Keywords: Microstates; 0301 basic medicine; Connectivity; MEG; Cognitive Neuroscience; Source reconstruction; Models, Neurological; Brain; Magnetoencephalography; Signal Processing, Computer-Assisted; Models, Theoretical; Hidden Markov Model; Article; 03 medical and health sciences; 0302 clinical medicine; Neurology; Humans; EEG

Our understanding of the dynamics neuronal activity in human brain remains limited, due part to a lack adequate methods for reconstructing from noninvasive electrophysiological data. Here, we present novel adaptive time-varying approach source reconstruction that can be applied magnetoencephalography (MEG) and electroencephalography (EEG) The method is underpinned by Hidden Markov Model (HMM), which infers points time when particular states re-occur sensor space HMM inference finds short-lived on scale 100ms. Intriguingly, this same timescale as EEG microstates. resulting state courses used intelligently pool data over these distinct periods time. This compute covariance matrices use beamforming, tune its spatial filtering properties those required at different Proof principle demonstrated with simulated data, demonstrate improvements MEG.