Beamformer Analysis of MEG Data

Magnetoencephalography (MEG) measures the magnetic fields outside the head created by electrical neuronal activity. The aim of many studies is to subsequently determine the spatiotemporal characteristics of these neuronal sources on the basis of the extracranial recordings, which means that an inverse problem needs to be solved. The MEG inverse problem is theoretically insoluble; just as inferring a three-dimensional scene from a two-dimensional image is insoluble. However, we are able to interpret cinematic images because we make certain assumptions about the world (the size of people, the way shadows fall) that allow us to achieve a percept. In MEG we are searching for a similar set of assumptions on which to base algorithms to interpret the MEG data. Recent work has shown that a class of algorithms used to solve the MEG inverse problem produce functionally plausible and verifiable results. These algorithms make the assumption that no two distinct cortical areas are perfectly linearly correlated in their activation time series and it has been shown empirically that this assumption is often justified. First, the spatial concurrence of beamformer images of induced neuronal activity and the BOLD (blood oxygenation level dependent) functional magnetic resonance imaging (fMRI) response was demonstrated in a biological motion and a letter fluency task (Singh et al., 2002) and more recently in a working memory task (Coppola et al., 2004). Second, beamformer analysis has been applied successfully in various experimental paradigms, ranging from experiments involving primary visual, auditory, and somatosensory cortices as well as the use of more cognitively demanding paradigms (e.g., Fawcett et al., 2004; Furlong et al., 2004; Gaetz and Cheyne, 2003; Hashimoto et al., 2001; Herdman et al., 2003; Hobson et al., 2005; Kamada et al., 1998; Ploner et al., 2002; Taniguchi et al., 2000; Ukai et al., 2002; also see Hillebrand et al., 2005, for review). One of the main advantages of beamformer analysis is that induced changes in cortical oscillatory power that do not result in a strong average-evoked response can be identified and localized. In particular, by using an active and control state, stimulus induced increases and decreases in cortical rhythms, known as event-related synchronization (ERS) and event-related desynchronization (ERD), respectively (Pfurtscheller and Lopes da Silva, 1999), can be quantified. Such changes in ongoing activity have been shown to play an important role in cognitive function (Arieli et al., 1996; Basar et al., 2001; Karakas et al., 2000; Kenet et al., 2003; Makeig et al., 2002; Ringach, 2003), and consequently form the basis of many theories of consciousness (Engel et al., 2001; Freeman, 2000; Llinás et al., 1998; Singer, 1998; Tononi and Edelman, 1998). Another advantage of beamformer analysis is that there is relatively little user interaction. The only parameters that a user needs to select are the size of the reconstruction grid, the time-frequency window over which to run the analysis, and optionally the amount of noise regularization. Importantly, there is no need to define the number of active sources a priori, since the beamformer output is computed for each voxel in the source-space independently and sequentially. The user friendliness of the technique makes it suitable for use in a clinical setting. This chapter is divided in two main sections. In the first section we describe the basic algorithmic steps that compose the beamformer and the characteristics of the reconstructed image of neuronal activity. We restrict almost all of our discussion to the measurement of induced and not evoked electrical activity. The beamformer approach has been used successfully in many experimental settings, hinting at the validity of the assumptions behind the technique. In the second section of this chapter we develop a case for why the beamformer assumption set, although simplistic, may indeed be quite plausible.