In this TRD project, we continue to develop methods for using noninvasive electromagnetic and hemodynamic data in conjunction with anatomical MRI for spatiotemporal brain imaging. Our new aims are directly motivated by new challenges arising from our collaborative research projects as discussed in our research strategy and in the descriptions of the collaborative projects. Magnetoencephalography and electroencephalography (MEG/EEG), constrained by structural and functional MRI, provide reliable recordings of cortical activity with millisecond precision. However, to date, recordings from subcortical structures, such as thalamus and the brainstem, have been limited due to low signal amplitudes and methodological challenges. To overcome these challenges in order to be able to reliably estimate the activity of several cortical regions and subcortical structures we will in our Aim 1 refine and extend our earlier work employing structured sparsity in mixed-norm source estimation (MxNE) approaches. Furthermore, it is of great interest to follow variation of brain activity on a trial-by-trial basis. Until now, such analyses have employed either linear distributed source models or dipole fitting methods, in which the source locations are determined on the basis of averaged data followed by trial-by-trial fitting of the source amplitudes. In our Aim 1, we will extend the MxNE methods to single-trial analysis of event-related data. A persistent problem in developing plausible neurophysiological models of perception, cognition and action is the difficulty of determining the how and when different neural systems interact. With the emerging interest in both anatomical and functional details of this brain connectivity, advanced methods to analyze electrophysiological, hemodynamic, and anatomical connectivity data, are of prime importance. To date, functional connectivity analysis has been often been limited to pairwise interactions. However, to properly account for the presence of multiple source regions in a network, a multivariate analysis is necessary. Therefore, in our Aim 2 we set out to develop novel methods, first applied to fMRI, data to assess connectivity in a network consisting of multiple nodes and in which each node comprises multiple voxels with time series inferred from fMRI data. At present, the electrophysiological basis of the robust resting-state networks observed in fMRI data is largely unknown. In our Aim 3, we will focus on methods that will facilitate making associations between MEG/EEG and fMRI networks. To this end, we will first develop techniques which consider the MEG/EEG source estimation problem as an integral part of the connectivity analysis and will develop methods tailored for estimation of ongoing activity, which has a stochastic nature as opposed to the deterministic signals modeled in event-related studies. Second, we will develop new approaches to relate hemodynamic and electromagnetic connectivity estimates. These advances will facilitate both neuroscientists and clinicians to make optimal use of their multimodal connectivity data.
Showing the most recent 10 out of 300 publications