Testing models of attention with MEG

Abstract

Neuroimaging methods have recently added new opportunities to test models of attention to deficit studies and approaches relying on invasive single and few unit electrophysiology. These techniques have very different time resolution so, a bridging technology is needed to link results from the slow hemodynamic methods and invasive electrophysiology. Electroencephalography (EEG) and magnetoencephalography (MEG) rely on non-invasive measures of mass electrical activity and are good candidates for this task. Some new insights have already been obtained using EEG and/or MEG in conjunction with fMRI. In these studies, the average EEG or MEG signal was used, often with constraints from fMRI, to study changes in the timecourse of regional activations under different attentional load. We postulate an engineering control framework that is broadly consistent with recent findings and generates predictions about when specific modules will become active under different attention paradigms. Specifically, the presence of relatively early temporal signals is predicted to occur in the parietal and frontal lobes as part of the attention-based CODAM model of consciousness. Tomographic analysis of single trial MEG data from a GO/NOGO task then identifies well-circumscribed regional increases and decreases of activity that confirm these general predictions. The results show how detailed information can be extracted from single subject data when the richness of information in the single trial MEG data is carefully analyzed and point the way for more elaborate analysis in the future.