The Added Value of EEG-fMRI in Imaging Neuroscience

The main objective of functional neuroimaging is to detect and characterise (in space and time) relevant changes in brain states and their relation to neuronal activity. Functional mri (fmri), electroencephalography (eeg) and magnetoencephalography (meg) are the most widespread noninvasive techniques that are available to experimental and clinical neuroscientists to achieve this objective starting from in vivo measures of brain electrical activity. Both fmri and eeg assume that a given brain state can be decoded from the precise anatomical localisation and the detailed temporal evolution of neuroelectrical brain activation signals, respectively. Starting from these common assumptions, fmri neuroscientists have developed many different approaches for mapping brain states at a spatial resolution of a few millimetres and testing many different neurophysiological and neuropathological hypotheses in normal and clinical populations, despite the limited temporal resolution of the available signals (see previous chapters). On the other hand, eeg neuroscientists have posed analogous questions and addressed similar problems by developing different approaches for the detailed temporal analysis of eeg recordings, despite the limited spatial detail in their findings. The previous chapter illustrated how fmri can be used by eeg neuroscientists to improve the quality of eeg results and to help with the problem of source localisation. The purpose of this chapter is to illustrate how the fmri neuroscientist can integrate detailed temporal information by incorporating simultaneously recorded eeg signals into standard as well as sophisticated fmri spatiotemporal modelling. We discuss how this can be achieved in such a way that new effects become detectable in the fmri domain even when the original event or state change causing possible fmri effects can only be characterised at very rapid temporal scales (e.g. Milliseconds) or frequency bands (above 1 hz). Our discussion occurs at a conceptual level, and we refer the reader to other chapters in part 2 for more details regarding problems such as eeg preprocessing.keywordsindependent component analysisfmri datafmri signalfmri experimentequivalent current dipolethese keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves