Magnetoencephalography (MEG) and Electroencephalography (EEG) are used to localise activity in the human brain, thereby providing the means to study its functional organization with a time resolution of 1 millisecond. Usually, the location of active regions is calculated relative to a sphere which is fitted to the outer surface of the head or to some known anatomical points. Structural information of the brain can be obtained with millimeter accuracy from Magnetic Resonance Images (MRI). The two types of information, i.e. structural and functional, can be combined using a coordinate system which is defined by anatomical landmarks on the head. Thus, active areas can be indicated within brain images, effectively giving an absolute location, within a certain error margin. This method is discussed in Chapter V. The resulting localization can be viewed in the original MRI slices, or in reconstructed three-dimensional views of the head or the brain. From these images cut-away views may be constructed, where the original MRI data is reconstructed in planes intersecting the location of the active region. The processing of the MRI scans to obtain these three-dimensional views is discussed in chapter IV.
In projecting the location of an active region of the brain on anatomical brain images with a resolution of one millimeter might suggest an accuracy in the location of the active region which is not realistic. An error estimate of the accuracy of the location is therefore necessary. To obtain such an estimate, it is important to consider the various aspects of the localization procedure which lead to an actual location. This is described in Chapter VII.
From the error estimate, and various simulations, it turns out that the noise in the MEG and EEG measurements has to be low enough for accurate localizations. Although averaging of evoked responses usually provides sufficiently high signal-to-noise ratios, various methods can be used to reduce noise levels even further during measurement. In Chapter II the performance of electronic balancing and of an active cancellation scheme is examined. As it turns out, the first method cannot be used within a magnetically shielded room. The second method turns out to be usable in the case that the measurements are performed within a magnetically shielded room.
Another aspect in the localization error is the large influence of the model used to represent the head as a volume conductor. Most commonly a multi-layered model, consisting of concentric spheres, is used. It is shown that results differ significantly if a realistically shaped compartment model is used. Such a model can be obtained from the MRI images. The processing of the MRI scan and the extraction of points at the various interfaces between regions of different conductivity, and the subsequent triangulation of these surfaces is discussed in chapter IV. The processing steps can be performed completely automatic, without human interference.
It is important to try to verify the localization results as much as possible. One of the methods is to compare the results obtained from MEG with those obtained from EEG. One other method is to compare the results with those obtained with Positron Emission Tomography (PET). Although it is not yet certain that localizations by means of MEG/EEG and PET should coincide, it is valuable to compare the results. A method to match PET and MRI is discussed in Chapter VI, together with an example combining four modalities, namely MEG, EEG, MRI and PET.
(c) MEG, EEG and the integration with Magnetic Resonance Images, H.J. Wieringa, 1993