Magnetoencephalography: fundamentals and established and emerging clinical applications in radiology

Abstract

Magnetoencephalography is a noninvasive, fast, and patient friendly technique for recording brain activity. It is increasingly available and is regarded as one of the most modern imaging tools available to radiologists. The dominant clinical use of this technology currently centers on two, partly overlapping areas, namely, localizing the regions from which epileptic seizures originate, and identifying regions of normal brain function in patients preparing to undergo brain surgery. As a consequence, many radiologists may not yet be familiar with this technique. This review provides an introduction to magnetoencephalography, discusses relevant analytical techniques, and presents recent developments in established and emerging clinical applications such as pervasive developmental disorders. Although the role of magnetoencephalography in diagnosis, prognosis, and patient treatment is still limited, it is argued that this technology is exquisitely capable of contributing indispensable information about brain dynamics not easily obtained with other modalities. This, it is believed, will make this technology an important clinical tool for a wide range of disorders in the future.

Figure 1

The MEG scanner at the Oxford Centre for Human Brain Activity. Left: Shown is a typical set up, where the subject is seated upright under the scanner. A button box is used to record behavioral responses and an eye-tracker supplies additional psychometrical data. Measurements in supine position are possible, and a doctor or nurse can stay within the magnetically shielded room in order to monitor the patient if needed. Middle: The brain magnetic fields are recorded with a helmet-shaped array of 102 SQUID devices featuring 3 pick-up coils each (306 channels in total). Two first-order gradiometers measure two orthogonal spatial gradients of the magnetic field in longitudinal and latitudinal directions, respectively. These channels are most sensitive to the tangential neuronal currents in the region below the device. A magnetometer coil (black rectangle) measures the magnetic field and is sensitive to deeper sources. Right: Typically, the cortex-to-detector distance is 4-5 cm. The primary output of the scanner is a trace for each channel representing the magnetic field or gradient as a function of time (for presentation, only 16 channels are shown; brain image courtesy of Elekta Neuromag OY, Helsinki).

Figure 2

Artifact removal. (a) Shown are data from one channel before (black) and after (red) ICA-based correction for eye-blink artifacts identified by EOG activity (bottom trace). The inset illustrates the 2-dimensional artifact space used for projection (time-courses and spatial topographies). The first dimension is dominant. Note the artifact has similar features as epileptic discharges in that it involves several MEG channels, has sharp peaks, and stands out from ongoing background activity (see also [55]). (b) Equivalent current dipole modeling of an evoked somatosensory response in a child before (left) and after (right) SSS-based movement correction. After correction, the ECD assumes a physiologically plausible location (courtesy of Elekta Neuromag OY, Helsinki). (c) Brain activity can be localized accurately despite strong artifacts caused by DBS electrodes (right). Before beamforming-based correction, strong, nonphysiological interferences outside the brain are observed (left, adapted from [56]).

Figure 3

MEG in presurgical evaluation. (a) Equivalent current dipole localization (center of circle) for an interictal discharge in a patient with right frontal epilepsy. (b) Distributed source modeling of language function in left hemisphere Wernicke's area for a verb generation task (adapted from [93]).

Figure 4

MEG in autism research. (a) The ECD locations for responses to images of human faces at about 145 ms after stimulus onset in a typically developing subject and an individual with ASD (circle indicates the volume conductor sphere). These images illustrate locations in the right posterior cortices of the generators, where, on average, dipole locations are more lateral in TD compared to ASD. (b) Grand root-mean-square signals following face images. The curves have been obtained by summation over all participants within a participant group (blue, boys with ASD; red, typically developing boys; and stimulus onset at 0) and channels. Even in middle childhood, the neural mechanisms underlying face processing are less specialized than in adults (inset) with greater early activation of posterior occipital cortices (I, II) and less specific activation of ventral occipitotemporal cortex (III), particularly in boys with ASD.