Imaging language function with MEG and fMRI

This Thesis considers the cortical mechanisms underlying language function, as measured by magnetoencephalography (MEG) and functional Magnetic Resonance Imaging (fMRI). In MEG, interpretation of the data critically depends on the ability to estimate the underlying neural activity and localize it to a certain part of the brain. In this Thesis, the accuracy of this procedure is explored in localization of cortical rhythms and for different head conductor models. A comparison of different source modeling techniques shows that rhythmic activity can be identified reliably with a variety of tools, such as equivalent current dipoles (ECDs), minimum norm approaches (MCEFD, minimum current estimates in the frequency domain), and beamformers (DICS, dynamic imaging of coherent sources). The results show that DICS is more sensitive to weak sources than the two other methods, both in measured and in simulated data. Computer simulations also demonstrate that for source localization performed under normal noisy conditions, a simple spherically symmetric head conductor model is in most cases a sufficient model for the conductivity geometry of the head. This Thesis specifically considers the cortical processing of action and object naming. We investigate whether different cortical regions are activated when actions or objects are named from the same images, and how the content of the image affects the brain correlates of naming. The MEG and fMRI results presented in this Thesis indicate that verbs and nouns are processed within the same cortical network, and demonstrate that image category (action/object) has a stronger influence than naming category (verb/noun) on the activation pattern within this network. In addition, we consider the relationship between MEG evoked responses and fMRI BOLD (blood-oxygen-level-dependent) signals in language tasks. We demonstrate differences between these two measures in both picture naming and reading, and show that such differences do not depend on experimental procedures such as different participants, languages, or task. In particular, we demonstrate an opposite stimulus effect for symbols and letter strings in the left occipito-temporal cortex in MEG vs. fMRI in reading, although the simultaneously measured electroencephalogram (EEG) was similar. We argue that the observed differences within this region reflect different neural generation mechanisms of the MEG evoked response and fMRI BOLD signals.