MEG measures the magnetic fields associated with brain activity using superconducting sensors placed around the head. The basis of the MEG signal is the macroscopic current flow in neural assemblies. If these neurons are aligned in parallel and fire synchronously the signal summates and becomes detectable over the ambient noise at the level of the sensory or electrode.
The distinct advantage of MEG is its high temporal resolution. The signal is directly related to neuronal activity, and the transmission of neuronal currents through the brain and to the sensors is virtually instantaneous, limited only by the sampling frequency of the recording equipment. This renders MEG ideally suited for testing hypotheses concerning the exact time course of brain processes.
The spatial resolution of MEG is somewhat restricted and can provide ambiguous answers. In dipole analysis, the local neuronal foci are usually modelled as equivalent current dipoles (ECD) whose number, strength and locations are estimated based on the externally measured magnetic field distribution. This procedure poses a non-trivial challenge because of the fact that there is no mathematically unique solution to the problem of inferring the numbers and locations of dipoles that could, theoretically, produce the observed pattern of activity on the surface of the skull, (i.e. there is an infinite number of source configurations that could produce exactly the same measured field). This is generally referred to as the inverse problem.
In practice, the experimenter uses a-priori knowledge of physiology and functional anatomy, often derived from other neuroimaging modalities, to incorporate feasible constraints into the model. For MEG data analysis we are using Dynamical Statistical Parametric Mapping (dSPM), a distributed, surface- constrained, noise-normalized source modeling technique. Besides restricting the source space to the surface, fMRI results from the same subject and task can be used to further bias the source solution.