Study of cortical rhythmic activity and connectivity with magnetoencephalography

Intracranial recordings in animals and neuroimaging studies on humans have indicated that oscillatory activity and its modulations may play a fundamental role in large-scale neural information processing. Furthermore, rhythmic interactions between cortical areas have been detected across a variety of tasks with electroencephalography (EEG) and magnetoencephalography (MEG). This kind of coupling has been proposed to be a key mechanism through which information is integrated across segregated areas. So far, rhythmic interactions have been analyzed primarily at the EEG/MEG sensor level, without explicit knowledge of cortical areas involved. In this thesis work we developed new methods that can be used to image oscillatory activity and coherence at the cortical level with MEG. Dynamic Imaging of Coherent Sources (DICS) enables localization of interacting areas both using external reference signals and directly from the MEG data. When the interacting areas have been determined it is possible to use additional measures beyond coherence to further quantify interactions within the networks. DICS was originally designed for study of continuous data; its further development into event-related DICS (erDICS) adds the possibility to image modulations of rhythmic activity that are locked to stimulus or movement timing. Furthermore, permutation testing incorporated into erDICS allows the evaluation of the statistical significance of the results. Analysis of simulated and real data showed that DICS and erDICS yield accurate localization and quantification of oscillatory activity and coherence. Comparison of DICS to other methods of localizing oscillatory activity revealed that it is equally accurate and that it can better separate the activity originating from two nearby areas. We applied DICS to two datasets, recorded from groups of subjects while they performed slow finger movements and when they were reading continuously. In both cases, we were able to systematically identify interacting cortico-cortical networks and, using phase coupling and causality measures, to quantify the manner in which the nodes within these networks influenced each other. Furthermore, we compared the identified reading network to results reported in neurophysiological and hemodynamic activation studies. In addition to areas typically detected in activation studies of reading the network included areas that are normally found in language production rather than perception tasks, indicating more extensive networking of neural systems than usually observed in activation studies