Finding a needle in the haystack: The challenges in using Magnetoencephalography to localize brain activity
Magnetoencephalography (MEG) is a non-invasive neuroimaging modality used to measure the magnetic fields associated with electrophysiological brain activity. This is accomplished by measuring the resulting magnetic fields outside the scalp surface that propagates from the currents generated by neuronal activity. These magnetic fields are very small with typical values between 10^-14 to 10^-13 Tesla. Detectability depends on the magnetic field’s relative magnitude compared to the effective background noise from both instrumentation and surroundings, which can be orders of magnitude larger than the target brain activity. Between 150 and 300 detectors are spatially distributed outside the scalp surface to measure the functioning brain at sub-millisecond resolution. These magnetic field brain data are analyzed at this sensor level or the underlying brain currents are reconstructed by inverting the equations relating the measured field to the brain currents. With large background noise, and a limited number of detector locations, significant challenges need to be overcome in order to effectively localize the target brain activity. However, the high temporal and frequency information contained in MEG data provides information about function and networks in typical and atypical brains that cannot be found in other neuroimaging modalities. The problem is underdetermined, non-unique and potentially non-Gaussian. I will describe some of the techniques used in MEG imaging that attempt to overcome some of these challenges and some of the applications that have emerged from the use of this technology.